All of the above --- fs/exec.c | 12 fs/lockd/host.c | 4 fs/locks.c | 48 - fs/namei.c | 95 +- fs/nfs/delegation.c | 6 fs/nfs/delegation.h | 16 fs/nfs/dir.c | 41 fs/nfs/file.c | 30 fs/nfs/inode.c | 105 +- fs/nfs/nfs3proc.c | 2 fs/nfs/nfs4_fs.h | 53 - fs/nfs/nfs4proc.c | 575 +++++++----- fs/nfs/nfs4state.c | 167 ++- fs/nfs/nfs4xdr.c | 62 - fs/nfs/proc.c | 2 fs/open.c | 79 + include/linux/namei.h | 8 include/linux/nfs_fs.h | 4 include/linux/nfs_xdr.h | 36 include/linux/sunrpc/debug.h | 3 include/linux/sunrpc/msg_prot.h | 25 include/linux/sunrpc/xdr.h | 6 include/linux/sunrpc/xprt.h | 224 +++- net/sunrpc/Makefile | 2 net/sunrpc/auth.c | 1 net/sunrpc/auth_gss/auth_gss.c | 8 net/sunrpc/auth_gss/gss_krb5_mech.c | 1 net/sunrpc/auth_gss/gss_mech_switch.c | 1 net/sunrpc/auth_null.c | 2 net/sunrpc/auth_unix.c | 2 net/sunrpc/clnt.c | 144 ++- net/sunrpc/pmap_clnt.c | 12 net/sunrpc/socklib.c | 175 +++ net/sunrpc/sunrpc_syms.c | 1 net/sunrpc/svcsock.c | 3 net/sunrpc/sysctl.c | 32 net/sunrpc/xdr.c | 177 --- net/sunrpc/xprt.c | 1610 +++++++++------------------------- net/sunrpc/xprtsock.c | 1252 ++++++++++++++++++++++++++ 39 files changed, 3072 insertions(+), 1954 deletions(-) Index: linux-2.6/fs/exec.c =================================================================== --- linux-2.6.orig/fs/exec.c +++ linux-2.6/fs/exec.c @@ -126,8 +126,7 @@ asmlinkage long sys_uselib(const char __ struct nameidata nd; int error; - nd.intent.open.flags = FMODE_READ; - error = __user_walk(library, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); + error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ); if (error) goto out; @@ -139,7 +138,7 @@ asmlinkage long sys_uselib(const char __ if (error) goto exit; - file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); + file = nameidata_to_filp(&nd, O_RDONLY); error = PTR_ERR(file); if (IS_ERR(file)) goto out; @@ -167,6 +166,7 @@ asmlinkage long sys_uselib(const char __ out: return error; exit: + release_open_intent(&nd); path_release(&nd); goto out; } @@ -490,8 +490,7 @@ struct file *open_exec(const char *name) int err; struct file *file; - nd.intent.open.flags = FMODE_READ; - err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_OPEN, &nd); + err = path_lookup_open(name, LOOKUP_FOLLOW, &nd, FMODE_READ); file = ERR_PTR(err); if (!err) { @@ -504,7 +503,7 @@ struct file *open_exec(const char *name) err = -EACCES; file = ERR_PTR(err); if (!err) { - file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); + file = nameidata_to_filp(&nd, O_RDONLY); if (!IS_ERR(file)) { err = deny_write_access(file); if (err) { @@ -516,6 +515,7 @@ out: return file; } } + release_open_intent(&nd); path_release(&nd); } goto out; Index: linux-2.6/fs/lockd/host.c =================================================================== --- linux-2.6.orig/fs/lockd/host.c +++ linux-2.6/fs/lockd/host.c @@ -173,11 +173,10 @@ nlm_bind_host(struct nlm_host *host) /* If we've already created an RPC client, check whether * RPC rebind is required - * Note: why keep rebinding if we're on a tcp connection? */ if ((clnt = host->h_rpcclnt) != NULL) { xprt = clnt->cl_xprt; - if (!xprt->stream && time_after_eq(jiffies, host->h_nextrebind)) { + if (time_after_eq(jiffies, host->h_nextrebind)) { clnt->cl_port = 0; host->h_nextrebind = jiffies + NLM_HOST_REBIND; dprintk("lockd: next rebind in %ld jiffies\n", @@ -189,7 +188,6 @@ nlm_bind_host(struct nlm_host *host) goto forgetit; xprt_set_timeout(&xprt->timeout, 5, nlmsvc_timeout); - xprt->nocong = 1; /* No congestion control for NLM */ xprt->resvport = 1; /* NLM requires a reserved port */ /* Existing NLM servers accept AUTH_UNIX only */ Index: linux-2.6/fs/locks.c =================================================================== --- linux-2.6.orig/fs/locks.c +++ linux-2.6/fs/locks.c @@ -316,21 +316,22 @@ static int flock_to_posix_lock(struct fi /* POSIX-1996 leaves the case l->l_len < 0 undefined; POSIX-2001 defines it. */ start += l->l_start; - end = start + l->l_len - 1; - if (l->l_len < 0) { + if (start < 0) + return -EINVAL; + fl->fl_end = OFFSET_MAX; + if (l->l_len > 0) { + end = start + l->l_len - 1; + fl->fl_end = end; + } else if (l->l_len < 0) { end = start - 1; + fl->fl_end = end; start += l->l_len; + if (start < 0) + return -EINVAL; } - - if (start < 0) - return -EINVAL; - if (l->l_len > 0 && end < 0) - return -EOVERFLOW; - fl->fl_start = start; /* we record the absolute position */ - fl->fl_end = end; - if (l->l_len == 0) - fl->fl_end = OFFSET_MAX; + if (fl->fl_end < fl->fl_start) + return -EOVERFLOW; fl->fl_owner = current->files; fl->fl_pid = current->tgid; @@ -362,14 +363,21 @@ static int flock64_to_posix_lock(struct return -EINVAL; } - if (((start += l->l_start) < 0) || (l->l_len < 0)) + start += l->l_start; + if (start < 0) return -EINVAL; - fl->fl_end = start + l->l_len - 1; - if (l->l_len > 0 && fl->fl_end < 0) - return -EOVERFLOW; + fl->fl_end = OFFSET_MAX; + if (l->l_len > 0) { + fl->fl_end = start + l->l_len - 1; + } else if (l->l_len < 0) { + fl->fl_end = start - 1; + start += l->l_len; + if (start < 0) + return -EINVAL; + } fl->fl_start = start; /* we record the absolute position */ - if (l->l_len == 0) - fl->fl_end = OFFSET_MAX; + if (fl->fl_end < fl->fl_start) + return -EOVERFLOW; fl->fl_owner = current->files; fl->fl_pid = current->tgid; @@ -829,12 +837,16 @@ static int __posix_lock_file(struct inod /* Detect adjacent or overlapping regions (if same lock type) */ if (request->fl_type == fl->fl_type) { + /* In all comparisons of start vs end, use + * "start - 1" rather than "end + 1". If end + * is OFFSET_MAX, end + 1 will become negative. + */ if (fl->fl_end < request->fl_start - 1) goto next_lock; /* If the next lock in the list has entirely bigger * addresses than the new one, insert the lock here. */ - if (fl->fl_start > request->fl_end + 1) + if (fl->fl_start - 1 > request->fl_end) break; /* If we come here, the new and old lock are of the Index: linux-2.6/fs/namei.c =================================================================== --- linux-2.6.orig/fs/namei.c +++ linux-2.6/fs/namei.c @@ -28,6 +28,7 @@ #include #include #include +#include #include #include @@ -317,6 +318,18 @@ void path_release_on_umount(struct namei mntput_no_expire(nd->mnt); } +/** + * release_open_intent - free up open intent resources + * @nd: pointer to nameidata + */ +void release_open_intent(struct nameidata *nd) +{ + if (nd->intent.open.file->f_dentry == NULL) + put_filp(nd->intent.open.file); + else + fput(nd->intent.open.file); +} + /* * Internal lookup() using the new generic dcache. * SMP-safe @@ -750,6 +763,7 @@ static fastcall int __link_path_walk(con struct qstr this; unsigned int c; + nd->flags |= LOOKUP_CONTINUE; err = exec_permission_lite(inode, nd); if (err == -EAGAIN) { err = permission(inode, MAY_EXEC, nd); @@ -802,7 +816,6 @@ static fastcall int __link_path_walk(con if (err < 0) break; } - nd->flags |= LOOKUP_CONTINUE; /* This does the actual lookups.. */ err = do_lookup(nd, &this, &next); if (err) @@ -1052,6 +1065,70 @@ out: return retval; } +static int __path_lookup_intent_open(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags, int create_mode) +{ + struct file *filp = get_empty_filp(); + int err; + + if (filp == NULL) + return -ENFILE; + nd->intent.open.file = filp; + nd->intent.open.flags = open_flags; + nd->intent.open.create_mode = create_mode; + err = path_lookup(name, lookup_flags|LOOKUP_OPEN, nd); + if (IS_ERR(nd->intent.open.file)) { + if (err == 0) { + err = PTR_ERR(nd->intent.open.file); + path_release(nd); + } + } else if (err != 0) + release_open_intent(nd); + return err; +} + +/** + * path_lookup_open - lookup a file path with open intent + * @name: pointer to file name + * @lookup_flags: lookup intent flags + * @nd: pointer to nameidata + * @open_flags: open intent flags + */ +int path_lookup_open(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags) +{ + return __path_lookup_intent_open(name, lookup_flags, nd, + open_flags, 0); +} + +/** + * path_lookup_create - lookup a file path with open + create intent + * @name: pointer to file name + * @lookup_flags: lookup intent flags + * @nd: pointer to nameidata + * @open_flags: open intent flags + * @create_mode: create intent flags + */ +int path_lookup_create(const char *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags, int create_mode) +{ + return __path_lookup_intent_open(name, lookup_flags|LOOKUP_CREATE, nd, + open_flags, create_mode); +} + +int __user_path_lookup_open(const char __user *name, unsigned int lookup_flags, + struct nameidata *nd, int open_flags) +{ + char *tmp = getname(name); + int err = PTR_ERR(tmp); + + if (!IS_ERR(tmp)) { + err = __path_lookup_intent_open(tmp, lookup_flags, nd, open_flags, 0); + putname(tmp); + } + return err; +} + /* * Restricted form of lookup. Doesn't follow links, single-component only, * needs parent already locked. Doesn't follow mounts. @@ -1416,27 +1493,27 @@ int may_open(struct nameidata *nd, int a */ int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) { - int acc_mode, error = 0; + int acc_mode, error; struct path path; struct dentry *dir; int count = 0; acc_mode = ACC_MODE(flag); + /* O_TRUNC implies we need access checks for write permissions */ + if (flag & O_TRUNC) + acc_mode |= MAY_WRITE; + /* Allow the LSM permission hook to distinguish append access from general write access. */ if (flag & O_APPEND) acc_mode |= MAY_APPEND; - /* Fill in the open() intent data */ - nd->intent.open.flags = flag; - nd->intent.open.create_mode = mode; - /* * The simplest case - just a plain lookup. */ if (!(flag & O_CREAT)) { - error = path_lookup(pathname, lookup_flags(flag)|LOOKUP_OPEN, nd); + error = path_lookup_open(pathname, lookup_flags(flag), nd, flag); if (error) return error; goto ok; @@ -1445,7 +1522,7 @@ int open_namei(const char * pathname, in /* * Create - we need to know the parent. */ - error = path_lookup(pathname, LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE, nd); + error = path_lookup_create(pathname, LOOKUP_PARENT, nd, flag, mode); if (error) return error; @@ -1520,6 +1597,8 @@ ok: exit_dput: dput_path(&path, nd); exit: + if (!IS_ERR(nd->intent.open.file)) + release_open_intent(nd); path_release(nd); return error; Index: linux-2.6/fs/nfs/delegation.c =================================================================== --- linux-2.6.orig/fs/nfs/delegation.c +++ linux-2.6/fs/nfs/delegation.c @@ -85,6 +85,10 @@ int nfs_inode_set_delegation(struct inod struct nfs_delegation *delegation; int status = 0; + /* Ensure we first revalidate the attributes and page cache! */ + if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR))) + __nfs_revalidate_inode(NFS_SERVER(inode), inode); + delegation = nfs_alloc_delegation(); if (delegation == NULL) return -ENOMEM; @@ -138,7 +142,7 @@ static void nfs_msync_inode(struct inode /* * Basic procedure for returning a delegation to the server */ -int nfs_inode_return_delegation(struct inode *inode) +int __nfs_inode_return_delegation(struct inode *inode) { struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state; struct nfs_inode *nfsi = NFS_I(inode); Index: linux-2.6/fs/nfs/delegation.h =================================================================== --- linux-2.6.orig/fs/nfs/delegation.h +++ linux-2.6/fs/nfs/delegation.h @@ -25,7 +25,7 @@ struct nfs_delegation { int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); -int nfs_inode_return_delegation(struct inode *inode); +int __nfs_inode_return_delegation(struct inode *inode); int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid); struct inode *nfs_delegation_find_inode(struct nfs4_client *clp, const struct nfs_fh *fhandle); @@ -47,11 +47,25 @@ static inline int nfs_have_delegation(st return 1; return 0; } + +static inline int nfs_inode_return_delegation(struct inode *inode) +{ + int err = 0; + + if (NFS_I(inode)->delegation != NULL) + err = __nfs_inode_return_delegation(inode); + return err; +} #else static inline int nfs_have_delegation(struct inode *inode, int flags) { return 0; } + +static inline int nfs_inode_return_delegation(struct inode *inode) +{ + return 0; +} #endif #endif Index: linux-2.6/fs/nfs/dir.c =================================================================== --- linux-2.6.orig/fs/nfs/dir.c +++ linux-2.6/fs/nfs/dir.c @@ -565,8 +565,6 @@ static int nfs_readdir(struct file *filp } } unlock_kernel(); - if (desc->error < 0) - return desc->error; if (res < 0) return res; return 0; @@ -803,6 +801,7 @@ static int nfs_dentry_delete(struct dent */ static void nfs_dentry_iput(struct dentry *dentry, struct inode *inode) { + nfs_inode_return_delegation(inode); if (dentry->d_flags & DCACHE_NFSFS_RENAMED) { lock_kernel(); inode->i_nlink--; @@ -916,7 +915,6 @@ static int is_atomic_open(struct inode * static struct dentry *nfs_atomic_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { struct dentry *res = NULL; - struct inode *inode = NULL; int error; /* Check that we are indeed trying to open this file */ @@ -930,8 +928,10 @@ static struct dentry *nfs_atomic_lookup( dentry->d_op = NFS_PROTO(dir)->dentry_ops; /* Let vfs_create() deal with O_EXCL */ - if (nd->intent.open.flags & O_EXCL) - goto no_entry; + if (nd->intent.open.flags & O_EXCL) { + d_add(dentry, NULL); + goto out; + } /* Open the file on the server */ lock_kernel(); @@ -945,32 +945,30 @@ static struct dentry *nfs_atomic_lookup( if (nd->intent.open.flags & O_CREAT) { nfs_begin_data_update(dir); - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); nfs_end_data_update(dir); } else - inode = nfs4_atomic_open(dir, dentry, nd); + res = nfs4_atomic_open(dir, dentry, nd); unlock_kernel(); - if (IS_ERR(inode)) { - error = PTR_ERR(inode); + if (IS_ERR(res)) { + error = PTR_ERR(res); switch (error) { /* Make a negative dentry */ case -ENOENT: - inode = NULL; - break; + res = NULL; + goto out; /* This turned out not to be a regular file */ + case -EISDIR: + case -ENOTDIR: + goto no_open; case -ELOOP: if (!(nd->intent.open.flags & O_NOFOLLOW)) goto no_open; - /* case -EISDIR: */ /* case -EINVAL: */ default: - res = ERR_PTR(error); goto out; } - } -no_entry: - res = d_add_unique(dentry, inode); - if (res != NULL) + } else if (res != NULL) dentry = res; nfs_renew_times(dentry); nfs_set_verifier(dentry, nfs_save_change_attribute(dir)); @@ -1014,7 +1012,7 @@ static int nfs_open_revalidate(struct de */ lock_kernel(); verifier = nfs_save_change_attribute(dir); - ret = nfs4_open_revalidate(dir, dentry, openflags); + ret = nfs4_open_revalidate(dir, dentry, openflags, nd); if (!ret) nfs_set_verifier(dentry, verifier); unlock_kernel(); @@ -1137,7 +1135,7 @@ static int nfs_create(struct inode *dir, lock_kernel(); nfs_begin_data_update(dir); - error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags); + error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags, nd); nfs_end_data_update(dir); if (error != 0) goto out_err; @@ -1332,6 +1330,7 @@ static int nfs_safe_remove(struct dentry nfs_begin_data_update(dir); if (inode != NULL) { + nfs_inode_return_delegation(inode); nfs_begin_data_update(inode); error = NFS_PROTO(dir)->remove(dir, &dentry->d_name); /* The VFS may want to delete this inode */ @@ -1539,7 +1538,8 @@ static int nfs_rename(struct inode *old_ #endif goto out; } - } + } else + new_inode->i_nlink--; go_ahead: /* @@ -1549,6 +1549,7 @@ go_ahead: nfs_wb_all(old_inode); shrink_dcache_parent(old_dentry); } + nfs_inode_return_delegation(old_inode); if (new_inode) d_delete(new_dentry); Index: linux-2.6/fs/nfs/file.c =================================================================== --- linux-2.6.orig/fs/nfs/file.c +++ linux-2.6/fs/nfs/file.c @@ -137,7 +137,8 @@ static int nfs_revalidate_file(struct in struct nfs_inode *nfsi = NFS_I(inode); int retval = 0; - if ((nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) || nfs_attribute_timeout(inode)) + if ((nfsi->cache_validity & (NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATTR)) + || nfs_attribute_timeout(inode)) retval = __nfs_revalidate_inode(NFS_SERVER(inode), inode); nfs_revalidate_mapping(inode, filp->f_mapping); return 0; @@ -375,22 +376,31 @@ out_swapfile: static int do_getlk(struct file *filp, int cmd, struct file_lock *fl) { + struct file_lock *cfl; struct inode *inode = filp->f_mapping->host; int status = 0; lock_kernel(); - /* Use local locking if mounted with "-onolock" */ - if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) - status = NFS_PROTO(inode)->lock(filp, cmd, fl); - else { - struct file_lock *cfl = posix_test_lock(filp, fl); - - fl->fl_type = F_UNLCK; - if (cfl != NULL) - memcpy(fl, cfl, sizeof(*fl)); + /* Try local locking first */ + cfl = posix_test_lock(filp, fl); + if (cfl != NULL) { + locks_copy_lock(fl, cfl); + goto out; } + + if (nfs_have_delegation(inode, FMODE_READ)) + goto out_noconflict; + + if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM) + goto out_noconflict; + + status = NFS_PROTO(inode)->lock(filp, cmd, fl); +out: unlock_kernel(); return status; +out_noconflict: + fl->fl_type = F_UNLCK; + goto out; } static int do_vfs_lock(struct file *file, struct file_lock *fl) Index: linux-2.6/fs/nfs/inode.c =================================================================== --- linux-2.6.orig/fs/nfs/inode.c +++ linux-2.6/fs/nfs/inode.c @@ -358,6 +358,35 @@ out_no_root: return no_root_error; } +static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans) +{ + to->to_initval = timeo * HZ / 10; + to->to_retries = retrans; + if (!to->to_retries) + to->to_retries = 2; + + switch (proto) { + case IPPROTO_TCP: + if (!to->to_initval) + to->to_initval = 60 * HZ; + if (to->to_initval > NFS_MAX_TCP_TIMEOUT) + to->to_initval = NFS_MAX_TCP_TIMEOUT; + to->to_increment = to->to_initval; + to->to_maxval = to->to_initval + (to->to_increment * to->to_retries); + to->to_exponential = 0; + break; + case IPPROTO_UDP: + default: + if (!to->to_initval) + to->to_initval = 11 * HZ / 10; + if (to->to_initval > NFS_MAX_UDP_TIMEOUT) + to->to_initval = NFS_MAX_UDP_TIMEOUT; + to->to_maxval = NFS_MAX_UDP_TIMEOUT; + to->to_exponential = 1; + break; + } +} + /* * Create an RPC client handle. */ @@ -367,22 +396,12 @@ nfs_create_client(struct nfs_server *ser struct rpc_timeout timeparms; struct rpc_xprt *xprt = NULL; struct rpc_clnt *clnt = NULL; - int tcp = (data->flags & NFS_MOUNT_TCP); + int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP; - /* Initialize timeout values */ - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT; - timeparms.to_exponential = 1; - - if (!timeparms.to_initval) - timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10; - if (!timeparms.to_retries) - timeparms.to_retries = 5; + nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans); /* create transport and client */ - xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP, - &server->addr, &timeparms); + xprt = xprt_create_proto(proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { dprintk("%s: cannot create RPC transport. Error = %ld\n", __FUNCTION__, PTR_ERR(xprt)); @@ -576,7 +595,6 @@ static int nfs_show_options(struct seq_f { NFS_MOUNT_SOFT, ",soft", ",hard" }, { NFS_MOUNT_INTR, ",intr", "" }, { NFS_MOUNT_POSIX, ",posix", "" }, - { NFS_MOUNT_TCP, ",tcp", ",udp" }, { NFS_MOUNT_NOCTO, ",nocto", "" }, { NFS_MOUNT_NOAC, ",noac", "" }, { NFS_MOUNT_NONLM, ",nolock", ",lock" }, @@ -585,6 +603,8 @@ static int nfs_show_options(struct seq_f }; struct proc_nfs_info *nfs_infop; struct nfs_server *nfss = NFS_SB(mnt->mnt_sb); + char buf[12]; + char *proto; seq_printf(m, ",v%d", nfss->rpc_ops->version); seq_printf(m, ",rsize=%d", nfss->rsize); @@ -603,6 +623,18 @@ static int nfs_show_options(struct seq_f else seq_puts(m, nfs_infop->nostr); } + switch (nfss->client->cl_xprt->prot) { + case IPPROTO_TCP: + proto = "tcp"; + break; + case IPPROTO_UDP: + proto = "udp"; + break; + default: + snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot); + proto = buf; + } + seq_printf(m, ",proto=%s", proto); seq_puts(m, ",addr="); seq_escape(m, nfss->hostname, " \t\n\\"); return 0; @@ -821,6 +853,11 @@ nfs_setattr(struct dentry *dentry, struc filemap_fdatawait(inode->i_mapping); nfs_wb_all(inode); } + /* + * Return any delegations if we're going to change ACLs + */ + if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) + nfs_inode_return_delegation(inode); error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr); if (error == 0) nfs_refresh_inode(inode, &fattr); @@ -877,12 +914,10 @@ static int nfs_wait_on_inode(struct inod sigset_t oldmask; int error; - atomic_inc(&inode->i_count); rpc_clnt_sigmask(clnt, &oldmask); error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING, nfs_wait_schedule, TASK_INTERRUPTIBLE); rpc_clnt_sigunmask(clnt, &oldmask); - iput(inode); return error; } @@ -1226,10 +1261,6 @@ int nfs_refresh_inode(struct inode *inod loff_t cur_size, new_isize; int data_unstable; - /* Do we hold a delegation? */ - if (nfs_have_delegation(inode, FMODE_READ)) - return 0; - spin_lock(&inode->i_lock); /* Are we in the process of updating data on the server? */ @@ -1350,7 +1381,8 @@ static int nfs_update_inode(struct inode nfsi->read_cache_jiffies = fattr->timestamp; /* Are we racing with known updates of the metadata on the server? */ - data_unstable = ! nfs_verify_change_attribute(inode, verifier); + data_unstable = ! (nfs_verify_change_attribute(inode, verifier) || + (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)); /* Check if our cached file size is stale */ new_isize = nfs_size_to_loff_t(fattr->size); @@ -1644,8 +1676,7 @@ static void nfs4_clear_inode(struct inod struct nfs_inode *nfsi = NFS_I(inode); /* If we are holding a delegation, return it! */ - if (nfsi->delegation != NULL) - nfs_inode_return_delegation(inode); + nfs_inode_return_delegation(inode); /* First call standard NFS clear_inode() code */ nfs_clear_inode(inode); /* Now clear out any remaining state */ @@ -1674,7 +1705,7 @@ static int nfs4_fill_super(struct super_ struct rpc_clnt *clnt = NULL; struct rpc_timeout timeparms; rpc_authflavor_t authflavour; - int proto, err = -EIO; + int err = -EIO; sb->s_blocksize_bits = 0; sb->s_blocksize = 0; @@ -1692,30 +1723,8 @@ static int nfs4_fill_super(struct super_ server->acdirmax = data->acdirmax*HZ; server->rpc_ops = &nfs_v4_clientops; - /* Initialize timeout values */ - - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_exponential = 1; - if (!timeparms.to_retries) - timeparms.to_retries = 5; - proto = data->proto; - /* Which IP protocol do we use? */ - switch (proto) { - case IPPROTO_TCP: - timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 600 * HZ / 10; - break; - case IPPROTO_UDP: - timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 11 * HZ / 10; - break; - default: - return -EINVAL; - } + nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans); clp = nfs4_get_client(&server->addr.sin_addr); if (!clp) { @@ -1740,7 +1749,7 @@ static int nfs4_fill_super(struct super_ down_write(&clp->cl_sem); if (IS_ERR(clp->cl_rpcclient)) { - xprt = xprt_create_proto(proto, &server->addr, &timeparms); + xprt = xprt_create_proto(data->proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { up_write(&clp->cl_sem); err = PTR_ERR(xprt); Index: linux-2.6/fs/nfs/nfs3proc.c =================================================================== --- linux-2.6.orig/fs/nfs/nfs3proc.c +++ linux-2.6/fs/nfs/nfs3proc.c @@ -299,7 +299,7 @@ static int nfs3_proc_commit(struct nfs_w */ static int nfs3_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs_fh fhandle; struct nfs_fattr fattr; Index: linux-2.6/fs/nfs/nfs4_fs.h =================================================================== --- linux-2.6.orig/fs/nfs/nfs4_fs.h +++ linux-2.6/fs/nfs/nfs4_fs.h @@ -93,25 +93,50 @@ struct nfs4_client { }; /* + * struct rpc_sequence ensures that RPC calls are sent in the exact + * order that they appear on the list. + */ +struct rpc_sequence { + struct rpc_wait_queue wait; /* RPC call delay queue */ + spinlock_t lock; /* Protects the list */ + struct list_head list; /* Defines sequence of RPC calls */ +}; + +#define NFS_SEQID_CONFIRMED 1 +struct nfs_seqid_counter { + struct rpc_sequence *sequence; + int flags; + u32 counter; +}; + +struct nfs_seqid { + struct list_head list; + struct nfs_seqid_counter *sequence; + struct rpc_task *task; +}; + +static inline void nfs_confirm_seqid(struct nfs_seqid_counter *seqid, int status) +{ + if (seqid_mutating_err(-status)) + seqid->flags |= NFS_SEQID_CONFIRMED; +} + +/* * NFS4 state_owners and lock_owners are simply labels for ordered * sequences of RPC calls. Their sole purpose is to provide once-only * semantics by allowing the server to identify replayed requests. - * - * The ->so_sema is held during all state_owner seqid-mutating operations: - * OPEN, OPEN_DOWNGRADE, and CLOSE. Its purpose is to properly serialize - * so_seqid. */ struct nfs4_state_owner { struct list_head so_list; /* per-clientid list of state_owners */ struct nfs4_client *so_client; u32 so_id; /* 32-bit identifier, unique */ - struct semaphore so_sema; - u32 so_seqid; /* protected by so_sema */ atomic_t so_count; struct rpc_cred *so_cred; /* Associated cred */ struct list_head so_states; struct list_head so_delegations; + struct nfs_seqid_counter so_seqid; + struct rpc_sequence so_sequence; }; /* @@ -132,7 +157,7 @@ struct nfs4_lock_state { fl_owner_t ls_owner; /* POSIX lock owner */ #define NFS_LOCK_INITIALIZED 1 int ls_flags; - u32 ls_seqid; + struct nfs_seqid_counter ls_seqid; u32 ls_id; nfs4_stateid ls_stateid; atomic_t ls_count; @@ -153,7 +178,6 @@ struct nfs4_state { struct inode *inode; /* Pointer to the inode */ unsigned long flags; /* Do we hold any locks? */ - struct semaphore lock_sema; /* Serializes file locking operations */ spinlock_t state_lock; /* Protects the lock_states list */ nfs4_stateid stateid; @@ -191,8 +215,8 @@ extern int nfs4_proc_setclientid_confirm extern int nfs4_proc_async_renew(struct nfs4_client *); extern int nfs4_proc_renew(struct nfs4_client *); extern int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode); -extern struct inode *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); -extern int nfs4_open_revalidate(struct inode *, struct dentry *, int); +extern struct dentry *nfs4_atomic_open(struct inode *, struct dentry *, struct nameidata *); +extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nameidata *); extern struct nfs4_state_recovery_ops nfs4_reboot_recovery_ops; extern struct nfs4_state_recovery_ops nfs4_network_partition_recovery_ops; @@ -224,12 +248,17 @@ extern struct nfs4_state * nfs4_get_open extern void nfs4_put_open_state(struct nfs4_state *); extern void nfs4_close_state(struct nfs4_state *, mode_t); extern struct nfs4_state *nfs4_find_state(struct inode *, struct rpc_cred *, mode_t mode); -extern void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp); extern void nfs4_schedule_state_recovery(struct nfs4_client *); +extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp); extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); -extern void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *ls); extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t); +extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter); +extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task); +extern void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid); +extern void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid); +extern void nfs_free_seqid(struct nfs_seqid *seqid); + extern const nfs4_stateid zero_stateid; /* nfs4xdr.c */ Index: linux-2.6/fs/nfs/nfs4proc.c =================================================================== --- linux-2.6.orig/fs/nfs/nfs4proc.c +++ linux-2.6/fs/nfs/nfs4proc.c @@ -47,6 +47,7 @@ #include #include #include +#include #include "nfs4_fs.h" #include "delegation.h" @@ -56,10 +57,11 @@ #define NFS4_POLL_RETRY_MIN (1*HZ) #define NFS4_POLL_RETRY_MAX (15*HZ) +static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid); static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *); -static int nfs4_async_handle_error(struct rpc_task *, struct nfs_server *); +static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *); static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry); -static int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception); +static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception); extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus); extern struct rpc_procinfo nfs4_procedures[]; @@ -189,6 +191,21 @@ static void update_changeattr(struct ino nfsi->change_attr = cinfo->after; } +/* Helper for asynchronous RPC calls */ +static int nfs4_call_async(struct rpc_clnt *clnt, rpc_action tk_begin, + rpc_action tk_exit, void *calldata) +{ + struct rpc_task *task; + + if (!(task = rpc_new_task(clnt, tk_exit, RPC_TASK_ASYNC))) + return -ENOMEM; + + task->tk_calldata = calldata; + task->tk_action = tk_begin; + rpc_execute(task); + return 0; +} + static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags) { struct inode *inode = state->inode; @@ -209,7 +226,6 @@ static void update_open_stateid(struct n /* * OPEN_RECLAIM: * reclaim state on the server after a reboot. - * Assumes caller is holding the sp->so_sem */ static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state) { @@ -218,7 +234,6 @@ static int _nfs4_open_reclaim(struct nfs struct nfs_delegation *delegation = NFS_I(inode)->delegation; struct nfs_openargs o_arg = { .fh = NFS_FH(inode), - .seqid = sp->so_seqid, .id = sp->so_id, .open_flags = state->state, .clientid = server->nfs4_state->cl_clientid, @@ -245,8 +260,13 @@ static int _nfs4_open_reclaim(struct nfs } o_arg.u.delegation_type = delegation->type; } + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + if (o_arg.seqid == NULL) + return -ENOMEM; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + /* Confirm the sequence as being established */ + nfs_confirm_seqid(&sp->so_seqid, status); + nfs_increment_open_seqid(status, o_arg.seqid); if (status == 0) { memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid)); if (o_res.delegation_type != 0) { @@ -256,6 +276,7 @@ static int _nfs4_open_reclaim(struct nfs nfs_async_inode_return_delegation(inode, &o_res.stateid); } } + nfs_free_seqid(o_arg.seqid); clear_bit(NFS_DELEGATED_STATE, &state->flags); /* Ensure we update the inode attributes */ NFS_CACHEINV(inode); @@ -302,23 +323,35 @@ static int _nfs4_open_delegation_recall( }; int status = 0; - down(&sp->so_sema); if (!test_bit(NFS_DELEGATED_STATE, &state->flags)) goto out; if (state->state == 0) goto out; - arg.seqid = sp->so_seqid; + arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + status = -ENOMEM; + if (arg.seqid == NULL) + goto out; arg.open_flags = state->state; memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data)); status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + nfs_increment_open_seqid(status, arg.seqid); + if (status != 0) + goto out_free; + if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) { + status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode), + sp, &res.stateid, arg.seqid); + if (status != 0) + goto out_free; + } + nfs_confirm_seqid(&sp->so_seqid, 0); if (status >= 0) { memcpy(state->stateid.data, res.stateid.data, sizeof(state->stateid.data)); clear_bit(NFS_DELEGATED_STATE, &state->flags); } +out_free: + nfs_free_seqid(arg.seqid); out: - up(&sp->so_sema); dput(parent); return status; } @@ -345,11 +378,11 @@ int nfs4_open_delegation_recall(struct d return err; } -static inline int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid) +static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid) { struct nfs_open_confirmargs arg = { .fh = fh, - .seqid = sp->so_seqid, + .seqid = seqid, .stateid = *stateid, }; struct nfs_open_confirmres res; @@ -362,7 +395,9 @@ static inline int _nfs4_proc_open_confir int status; status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + /* Confirm the sequence as being established */ + nfs_confirm_seqid(&sp->so_seqid, status); + nfs_increment_open_seqid(status, seqid); if (status >= 0) memcpy(stateid, &res.stateid, sizeof(*stateid)); return status; @@ -380,21 +415,37 @@ static int _nfs4_proc_open(struct inode int status; /* Update sequence id. The caller must serialize! */ - o_arg->seqid = sp->so_seqid; o_arg->id = sp->so_id; o_arg->clientid = sp->so_client->cl_clientid; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_seqid(status, sp); + if (status == 0) { + /* OPEN on anything except a regular file is disallowed in NFSv4 */ + switch (o_res->f_attr->mode & S_IFMT) { + case S_IFREG: + break; + case S_IFLNK: + status = -ELOOP; + break; + case S_IFDIR: + status = -EISDIR; + break; + default: + status = -ENOTDIR; + } + } + + nfs_increment_open_seqid(status, o_arg->seqid); if (status != 0) goto out; update_changeattr(dir, &o_res->cinfo); if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) { status = _nfs4_proc_open_confirm(server->client, &o_res->fh, - sp, &o_res->stateid); + sp, &o_res->stateid, o_arg->seqid); if (status != 0) goto out; } + nfs_confirm_seqid(&sp->so_seqid, 0); if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr); out: @@ -465,6 +516,10 @@ static int _nfs4_open_expired(struct nfs set_bit(NFS_DELEGATED_STATE, &state->flags); goto out; } + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + status = -ENOMEM; + if (o_arg.seqid == NULL) + goto out; status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_nodeleg; @@ -490,6 +545,7 @@ static int _nfs4_open_expired(struct nfs nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res); } out_nodeleg: + nfs_free_seqid(o_arg.seqid); clear_bit(NFS_DELEGATED_STATE, &state->flags); out: dput(parent); @@ -564,7 +620,6 @@ static int _nfs4_open_delegated(struct i dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__); goto out_err; } - down(&sp->so_sema); state = nfs4_get_open_state(inode, sp); if (state == NULL) goto out_err; @@ -589,7 +644,6 @@ static int _nfs4_open_delegated(struct i set_bit(NFS_DELEGATED_STATE, &state->flags); update_open_stateid(state, &delegation->stateid, open_flags); out_ok: - up(&sp->so_sema); nfs4_put_state_owner(sp); up_read(&nfsi->rwsem); up_read(&clp->cl_sem); @@ -600,11 +654,12 @@ out_err: if (sp != NULL) { if (state != NULL) nfs4_put_open_state(state); - up(&sp->so_sema); nfs4_put_state_owner(sp); } up_read(&nfsi->rwsem); up_read(&clp->cl_sem); + if (err != -EACCES) + nfs_inode_return_delegation(inode); return err; } @@ -665,8 +720,10 @@ static int _nfs4_do_open(struct inode *d } else o_arg.u.attrs = sattr; /* Serialization for the sequence id */ - down(&sp->so_sema); + o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid); + if (o_arg.seqid == NULL) + return -ENOMEM; status = _nfs4_proc_open(dir, sp, &o_arg, &o_res); if (status != 0) goto out_err; @@ -681,7 +738,7 @@ static int _nfs4_do_open(struct inode *d update_open_stateid(state, &o_res.stateid, flags); if (o_res.delegation_type != 0) nfs_inode_set_delegation(inode, cred, &o_res); - up(&sp->so_sema); + nfs_free_seqid(o_arg.seqid); nfs4_put_state_owner(sp); up_read(&clp->cl_sem); *res = state; @@ -690,7 +747,7 @@ out_err: if (sp != NULL) { if (state != NULL) nfs4_put_open_state(state); - up(&sp->so_sema); + nfs_free_seqid(o_arg.seqid); nfs4_put_state_owner(sp); } /* Note: clp->cl_sem must be released before nfs4_put_open_state()! */ @@ -718,7 +775,7 @@ static struct nfs4_state *nfs4_do_open(s * It is actually a sign of a bug on the client or on the server. * * If we receive a BAD_SEQID error in the particular case of - * doing an OPEN, we assume that nfs4_increment_seqid() will + * doing an OPEN, we assume that nfs_increment_open_seqid() will * have unhashed the old state_owner for us, and that we can * therefore safely retry using a new one. We should still warn * the user though... @@ -728,6 +785,16 @@ static struct nfs4_state *nfs4_do_open(s exception.retry = 1; continue; } + /* + * BAD_STATEID on OPEN means that the server cancelled our + * state before it received the OPEN_CONFIRM. + * Recover by retrying the request as per the discussion + * on Page 181 of RFC3530. + */ + if (status == -NFS4ERR_BAD_STATEID) { + exception.retry = 1; + continue; + } res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir), status, &exception)); } while (exception.retry); @@ -789,17 +856,27 @@ struct nfs4_closedata { struct nfs_closeres res; }; +static void nfs4_free_closedata(struct nfs4_closedata *calldata) +{ + struct nfs4_state *state = calldata->state; + struct nfs4_state_owner *sp = state->owner; + + nfs4_put_open_state(calldata->state); + nfs_free_seqid(calldata->arg.seqid); + nfs4_put_state_owner(sp); + kfree(calldata); +} + static void nfs4_close_done(struct rpc_task *task) { struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; struct nfs4_state *state = calldata->state; - struct nfs4_state_owner *sp = state->owner; struct nfs_server *server = NFS_SERVER(calldata->inode); /* hmm. we are done with the inode, and in the process of freeing * the state_owner. we keep this around to process errors */ - nfs4_increment_seqid(task->tk_status, sp); + nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid); switch (task->tk_status) { case 0: memcpy(&state->stateid, &calldata->res.stateid, @@ -817,24 +894,46 @@ static void nfs4_close_done(struct rpc_t } } state->state = calldata->arg.open_flags; - nfs4_put_open_state(state); - up(&sp->so_sema); - nfs4_put_state_owner(sp); - up_read(&server->nfs4_state->cl_sem); - kfree(calldata); + nfs4_free_closedata(calldata); } -static inline int nfs4_close_call(struct rpc_clnt *clnt, struct nfs4_closedata *calldata) +static void nfs4_close_begin(struct rpc_task *task) { + struct nfs4_closedata *calldata = (struct nfs4_closedata *)task->tk_calldata; + struct nfs4_state *state = calldata->state; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE], .rpc_argp = &calldata->arg, .rpc_resp = &calldata->res, - .rpc_cred = calldata->state->owner->so_cred, + .rpc_cred = state->owner->so_cred, }; - if (calldata->arg.open_flags != 0) + int mode = 0; + int status; + + status = nfs_wait_on_sequence(calldata->arg.seqid, task); + if (status != 0) + return; + /* Don't reorder reads */ + smp_rmb(); + /* Recalculate the new open mode in case someone reopened the file + * while we were waiting in line to be scheduled. + */ + if (state->nreaders != 0) + mode |= FMODE_READ; + if (state->nwriters != 0) + mode |= FMODE_WRITE; + if (test_bit(NFS_DELEGATED_STATE, &state->flags)) + state->state = mode; + if (mode == state->state) { + nfs4_free_closedata(calldata); + task->tk_exit = NULL; + rpc_exit(task, 0); + return; + } + if (mode != 0) msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE]; - return rpc_call_async(clnt, &msg, 0, nfs4_close_done, calldata); + calldata->arg.open_flags = mode; + rpc_call_setup(task, &msg, 0); } /* @@ -851,39 +950,52 @@ static inline int nfs4_close_call(struct int nfs4_do_close(struct inode *inode, struct nfs4_state *state, mode_t mode) { struct nfs4_closedata *calldata; - int status; + int status = -ENOMEM; - /* Tell caller we're done */ - if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { - state->state = mode; - return 0; - } - calldata = (struct nfs4_closedata *)kmalloc(sizeof(*calldata), GFP_KERNEL); + calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); if (calldata == NULL) - return -ENOMEM; + goto out; calldata->inode = inode; calldata->state = state; calldata->arg.fh = NFS_FH(inode); + calldata->arg.stateid = &state->stateid; /* Serialization for the sequence id */ - calldata->arg.seqid = state->owner->so_seqid; - calldata->arg.open_flags = mode; - memcpy(&calldata->arg.stateid, &state->stateid, - sizeof(calldata->arg.stateid)); - status = nfs4_close_call(NFS_SERVER(inode)->client, calldata); - /* - * Return -EINPROGRESS on success in order to indicate to the - * caller that an asynchronous RPC call has been launched, and - * that it will release the semaphores on completion. - */ - return (status == 0) ? -EINPROGRESS : status; + calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid); + if (calldata->arg.seqid == NULL) + goto out_free_calldata; + + status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_close_begin, + nfs4_close_done, calldata); + if (status == 0) + goto out; + + nfs_free_seqid(calldata->arg.seqid); +out_free_calldata: + kfree(calldata); +out: + return status; +} + +static void nfs4_intent_set_file(struct nameidata *nd, struct dentry *dentry, struct nfs4_state *state) +{ + struct file *filp; + + filp = lookup_instantiate_filp(nd, dentry, NULL); + if (!IS_ERR(filp)) { + struct nfs_open_context *ctx; + ctx = (struct nfs_open_context *)filp->private_data; + ctx->state = state; + } else + nfs4_close_state(state, nd->intent.open.flags); } -struct inode * +struct dentry * nfs4_atomic_open(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { struct iattr attr; struct rpc_cred *cred; struct nfs4_state *state; + struct dentry *res; if (nd->flags & LOOKUP_CREATE) { attr.ia_mode = nd->intent.open.create_mode; @@ -897,16 +1009,23 @@ nfs4_atomic_open(struct inode *dir, stru cred = rpcauth_lookupcred(NFS_SERVER(dir)->client->cl_auth, 0); if (IS_ERR(cred)) - return (struct inode *)cred; + return (struct dentry *)cred; state = nfs4_do_open(dir, dentry, nd->intent.open.flags, &attr, cred); put_rpccred(cred); - if (IS_ERR(state)) - return (struct inode *)state; - return state->inode; + if (IS_ERR(state)) { + if (PTR_ERR(state) == -ENOENT) + d_add(dentry, NULL); + return (struct dentry *)state; + } + res = d_add_unique(dentry, state->inode); + if (res != NULL) + dentry = res; + nfs4_intent_set_file(nd, dentry, state); + return res; } int -nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags) +nfs4_open_revalidate(struct inode *dir, struct dentry *dentry, int openflags, struct nameidata *nd) { struct rpc_cred *cred; struct nfs4_state *state; @@ -919,18 +1038,30 @@ nfs4_open_revalidate(struct inode *dir, if (IS_ERR(state)) state = nfs4_do_open(dir, dentry, openflags, NULL, cred); put_rpccred(cred); - if (state == ERR_PTR(-ENOENT) && dentry->d_inode == 0) - return 1; - if (IS_ERR(state)) - return 0; + if (IS_ERR(state)) { + switch (PTR_ERR(state)) { + case -EPERM: + case -EACCES: + case -EDQUOT: + case -ENOSPC: + case -EROFS: + lookup_instantiate_filp(nd, (struct dentry *)state, NULL); + return 1; + case -ENOENT: + if (dentry->d_inode == NULL) + return 1; + } + goto out_drop; + } inode = state->inode; + iput(inode); if (inode == dentry->d_inode) { - iput(inode); + nfs4_intent_set_file(nd, dentry, state); return 1; } - d_drop(dentry); nfs4_close_state(state, openflags); - iput(inode); +out_drop: + d_drop(dentry); return 0; } @@ -1431,7 +1562,7 @@ static int nfs4_proc_commit(struct nfs_w static int nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs4_state *state; struct rpc_cred *cred; @@ -1453,13 +1584,13 @@ nfs4_proc_create(struct inode *dir, stru struct nfs_fattr fattr; status = nfs4_do_setattr(NFS_SERVER(dir), &fattr, NFS_FH(state->inode), sattr, state); - if (status == 0) { + if (status == 0) nfs_setattr_update_inode(state->inode, sattr); - goto out; - } - } else if (flags != 0) - goto out; - nfs4_close_state(state, flags); + } + if (status == 0 && nd != NULL && (nd->flags & LOOKUP_OPEN)) + nfs4_intent_set_file(nd, dentry, state); + else + nfs4_close_state(state, flags); out: return status; } @@ -2106,65 +2237,6 @@ nfs4_proc_renew(struct nfs4_client *clp) return 0; } -/* - * We will need to arrange for the VFS layer to provide an atomic open. - * Until then, this open method is prone to inefficiency and race conditions - * due to the lookup, potential create, and open VFS calls from sys_open() - * placed on the wire. - */ -static int -nfs4_proc_file_open(struct inode *inode, struct file *filp) -{ - struct dentry *dentry = filp->f_dentry; - struct nfs_open_context *ctx; - struct nfs4_state *state = NULL; - struct rpc_cred *cred; - int status = -ENOMEM; - - dprintk("nfs4_proc_file_open: starting on (%.*s/%.*s)\n", - (int)dentry->d_parent->d_name.len, - dentry->d_parent->d_name.name, - (int)dentry->d_name.len, dentry->d_name.name); - - - /* Find our open stateid */ - cred = rpcauth_lookupcred(NFS_SERVER(inode)->client->cl_auth, 0); - if (IS_ERR(cred)) - return PTR_ERR(cred); - ctx = alloc_nfs_open_context(dentry, cred); - put_rpccred(cred); - if (unlikely(ctx == NULL)) - return -ENOMEM; - status = -EIO; /* ERACE actually */ - state = nfs4_find_state(inode, cred, filp->f_mode); - if (unlikely(state == NULL)) - goto no_state; - ctx->state = state; - nfs4_close_state(state, filp->f_mode); - ctx->mode = filp->f_mode; - nfs_file_set_open_context(filp, ctx); - put_nfs_open_context(ctx); - if (filp->f_mode & FMODE_WRITE) - nfs_begin_data_update(inode); - return 0; -no_state: - printk(KERN_WARNING "NFS: v4 raced in function %s\n", __FUNCTION__); - put_nfs_open_context(ctx); - return status; -} - -/* - * Release our state - */ -static int -nfs4_proc_file_release(struct inode *inode, struct file *filp) -{ - if (filp->f_mode & FMODE_WRITE) - nfs_end_data_update(inode); - nfs_file_clear_open_context(filp); - return 0; -} - static inline int nfs4_server_supports_acls(struct nfs_server *server) { return (server->caps & NFS_CAP_ACLS) @@ -2345,6 +2417,7 @@ static int nfs4_proc_set_acl(struct inod if (!nfs4_server_supports_acls(server)) return -EOPNOTSUPP; + nfs_inode_return_delegation(inode); buf_to_pages(buf, buflen, arg.acl_pages, &arg.acl_pgbase); ret = rpc_call_sync(NFS_SERVER(inode)->client, &msg, 0); if (ret == 0) @@ -2353,7 +2426,7 @@ static int nfs4_proc_set_acl(struct inod } static int -nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server) +nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server) { struct nfs4_client *clp = server->nfs4_state; @@ -2431,7 +2504,7 @@ static int nfs4_delay(struct rpc_clnt *c /* This is the error handling routine for processes that are allowed * to sleep. */ -int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception) +int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception) { struct nfs4_client *clp = server->nfs4_state; int ret = errorcode; @@ -2632,7 +2705,6 @@ static int _nfs4_proc_getlk(struct nfs4_ down_read(&clp->cl_sem); nlo.clientid = clp->cl_clientid; - down(&state->lock_sema); status = nfs4_set_lock_state(state, request); if (status != 0) goto out; @@ -2659,7 +2731,6 @@ static int _nfs4_proc_getlk(struct nfs4_ status = 0; } out: - up(&state->lock_sema); up_read(&clp->cl_sem); return status; } @@ -2696,67 +2767,125 @@ static int do_vfs_lock(struct file *file return res; } -static int _nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) +struct nfs4_unlockdata { + struct nfs_lockargs arg; + struct nfs_locku_opargs luargs; + struct nfs_lockres res; + struct nfs4_lock_state *lsp; + struct nfs_open_context *ctx; + atomic_t refcount; + struct completion completion; +}; + +static void nfs4_locku_release_calldata(struct nfs4_unlockdata *calldata) { - struct inode *inode = state->inode; - struct nfs_server *server = NFS_SERVER(inode); - struct nfs4_client *clp = server->nfs4_state; - struct nfs_lockargs arg = { - .fh = NFS_FH(inode), - .type = nfs4_lck_type(cmd, request), - .offset = request->fl_start, - .length = nfs4_lck_length(request), - }; - struct nfs_lockres res = { - .server = server, - }; + if (atomic_dec_and_test(&calldata->refcount)) { + nfs_free_seqid(calldata->luargs.seqid); + nfs4_put_lock_state(calldata->lsp); + put_nfs_open_context(calldata->ctx); + kfree(calldata); + } +} + +static void nfs4_locku_complete(struct nfs4_unlockdata *calldata) +{ + complete(&calldata->completion); + nfs4_locku_release_calldata(calldata); +} + +static void nfs4_locku_done(struct rpc_task *task) +{ + struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; + + nfs_increment_lock_seqid(task->tk_status, calldata->luargs.seqid); + switch (task->tk_status) { + case 0: + memcpy(calldata->lsp->ls_stateid.data, + calldata->res.u.stateid.data, + sizeof(calldata->lsp->ls_stateid.data)); + break; + case -NFS4ERR_STALE_STATEID: + case -NFS4ERR_EXPIRED: + nfs4_schedule_state_recovery(calldata->res.server->nfs4_state); + break; + default: + if (nfs4_async_handle_error(task, calldata->res.server) == -EAGAIN) { + rpc_restart_call(task); + return; + } + } + nfs4_locku_complete(calldata); +} + +static void nfs4_locku_begin(struct rpc_task *task) +{ + struct nfs4_unlockdata *calldata = (struct nfs4_unlockdata *)task->tk_calldata; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU], - .rpc_argp = &arg, - .rpc_resp = &res, - .rpc_cred = state->owner->so_cred, + .rpc_argp = &calldata->arg, + .rpc_resp = &calldata->res, + .rpc_cred = calldata->lsp->ls_state->owner->so_cred, }; + int status; + + status = nfs_wait_on_sequence(calldata->luargs.seqid, task); + if (status != 0) + return; + if ((calldata->lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) { + nfs4_locku_complete(calldata); + task->tk_exit = NULL; + rpc_exit(task, 0); + return; + } + rpc_call_setup(task, &msg, 0); +} + +static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) +{ + struct nfs4_unlockdata *calldata; + struct inode *inode = state->inode; + struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp; - struct nfs_locku_opargs luargs; int status; - - down_read(&clp->cl_sem); - down(&state->lock_sema); + status = nfs4_set_lock_state(state, request); if (status != 0) - goto out; + return status; lsp = request->fl_u.nfs4_fl.owner; /* We might have lost the locks! */ if ((lsp->ls_flags & NFS_LOCK_INITIALIZED) == 0) - goto out; - luargs.seqid = lsp->ls_seqid; - memcpy(&luargs.stateid, &lsp->ls_stateid, sizeof(luargs.stateid)); - arg.u.locku = &luargs; - status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - nfs4_increment_lock_seqid(status, lsp); + return 0; + calldata = kmalloc(sizeof(*calldata), GFP_KERNEL); + if (calldata == NULL) + return -ENOMEM; + calldata->luargs.seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (calldata->luargs.seqid == NULL) { + kfree(calldata); + return -ENOMEM; + } + calldata->luargs.stateid = &lsp->ls_stateid; + calldata->arg.fh = NFS_FH(inode); + calldata->arg.type = nfs4_lck_type(cmd, request); + calldata->arg.offset = request->fl_start; + calldata->arg.length = nfs4_lck_length(request); + calldata->arg.u.locku = &calldata->luargs; + calldata->res.server = server; + calldata->lsp = lsp; + atomic_inc(&lsp->ls_count); - if (status == 0) - memcpy(&lsp->ls_stateid, &res.u.stateid, - sizeof(lsp->ls_stateid)); -out: - up(&state->lock_sema); - if (status == 0) - do_vfs_lock(request->fl_file, request); - up_read(&clp->cl_sem); - return status; -} + /* Ensure we don't close file until we're done freeing locks! */ + calldata->ctx = get_nfs_open_context((struct nfs_open_context*)request->fl_file->private_data); -static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request) -{ - struct nfs4_exception exception = { }; - int err; + atomic_set(&calldata->refcount, 2); + init_completion(&calldata->completion); - do { - err = nfs4_handle_exception(NFS_SERVER(state->inode), - _nfs4_proc_unlck(state, cmd, request), - &exception); - } while (exception.retry); - return err; + status = nfs4_call_async(NFS_SERVER(inode)->client, nfs4_locku_begin, + nfs4_locku_done, calldata); + if (status == 0) + wait_for_completion_interruptible(&calldata->completion); + do_vfs_lock(request->fl_file, request); + nfs4_locku_release_calldata(calldata); + return status; } static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim) @@ -2764,11 +2893,23 @@ static int _nfs4_do_setlk(struct nfs4_st struct inode *inode = state->inode; struct nfs_server *server = NFS_SERVER(inode); struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner; + struct nfs_lock_opargs largs = { + .lock_stateid = &lsp->ls_stateid, + .open_stateid = &state->stateid, + .lock_owner = { + .clientid = server->nfs4_state->cl_clientid, + .id = lsp->ls_id, + }, + .reclaim = reclaim, + }; struct nfs_lockargs arg = { .fh = NFS_FH(inode), .type = nfs4_lck_type(cmd, request), .offset = request->fl_start, .length = nfs4_lck_length(request), + .u = { + .lock = &largs, + }, }; struct nfs_lockres res = { .server = server, @@ -2779,53 +2920,39 @@ static int _nfs4_do_setlk(struct nfs4_st .rpc_resp = &res, .rpc_cred = state->owner->so_cred, }; - struct nfs_lock_opargs largs = { - .reclaim = reclaim, - .new_lock_owner = 0, - }; - int status; + int status = -ENOMEM; - if (!(lsp->ls_flags & NFS_LOCK_INITIALIZED)) { + largs.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid); + if (largs.lock_seqid == NULL) + return -ENOMEM; + if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) { struct nfs4_state_owner *owner = state->owner; - struct nfs_open_to_lock otl = { - .lock_owner = { - .clientid = server->nfs4_state->cl_clientid, - }, - }; - - otl.lock_seqid = lsp->ls_seqid; - otl.lock_owner.id = lsp->ls_id; - memcpy(&otl.open_stateid, &state->stateid, sizeof(otl.open_stateid)); - largs.u.open_lock = &otl; + + largs.open_seqid = nfs_alloc_seqid(&owner->so_seqid); + if (largs.open_seqid == NULL) + goto out; largs.new_lock_owner = 1; - arg.u.lock = &largs; - down(&owner->so_sema); - otl.open_seqid = owner->so_seqid; status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment open_owner seqid on success, and - * seqid mutating errors */ - nfs4_increment_seqid(status, owner); - up(&owner->so_sema); - if (status == 0) { - lsp->ls_flags |= NFS_LOCK_INITIALIZED; - lsp->ls_seqid++; + /* increment open seqid on success, and seqid mutating errors */ + if (largs.new_lock_owner != 0) { + nfs_increment_open_seqid(status, largs.open_seqid); + if (status == 0) + nfs_confirm_seqid(&lsp->ls_seqid, 0); } - } else { - struct nfs_exist_lock el = { - .seqid = lsp->ls_seqid, - }; - memcpy(&el.stateid, &lsp->ls_stateid, sizeof(el.stateid)); - largs.u.exist_lock = ⪙ - arg.u.lock = &largs; + nfs_free_seqid(largs.open_seqid); + } else status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR); - /* increment seqid on success, and * seqid mutating errors*/ - nfs4_increment_lock_seqid(status, lsp); - } + /* increment lock seqid on success, and seqid mutating errors*/ + nfs_increment_lock_seqid(status, largs.lock_seqid); /* save the returned stateid. */ - if (status == 0) - memcpy(&lsp->ls_stateid, &res.u.stateid, sizeof(nfs4_stateid)); - else if (status == -NFS4ERR_DENIED) + if (status == 0) { + memcpy(lsp->ls_stateid.data, res.u.stateid.data, + sizeof(lsp->ls_stateid.data)); + lsp->ls_flags |= NFS_LOCK_INITIALIZED; + } else if (status == -NFS4ERR_DENIED) status = -EAGAIN; +out: + nfs_free_seqid(largs.lock_seqid); return status; } @@ -2865,11 +2992,9 @@ static int _nfs4_proc_setlk(struct nfs4_ int status; down_read(&clp->cl_sem); - down(&state->lock_sema); status = nfs4_set_lock_state(state, request); if (status == 0) status = _nfs4_do_setlk(state, cmd, request, 0); - up(&state->lock_sema); if (status == 0) { /* Note: we always want to sleep here! */ request->fl_flags |= FL_SLEEP; @@ -3024,8 +3149,8 @@ struct nfs_rpc_ops nfs_v4_clientops = { .read_setup = nfs4_proc_read_setup, .write_setup = nfs4_proc_write_setup, .commit_setup = nfs4_proc_commit_setup, - .file_open = nfs4_proc_file_open, - .file_release = nfs4_proc_file_release, + .file_open = nfs_open, + .file_release = nfs_release, .lock = nfs4_proc_lock, .clear_acl_cache = nfs4_zap_acl_attr, }; Index: linux-2.6/fs/nfs/nfs4state.c =================================================================== --- linux-2.6.orig/fs/nfs/nfs4state.c +++ linux-2.6/fs/nfs/nfs4state.c @@ -264,13 +264,15 @@ nfs4_alloc_state_owner(void) { struct nfs4_state_owner *sp; - sp = kmalloc(sizeof(*sp),GFP_KERNEL); + sp = kzalloc(sizeof(*sp),GFP_KERNEL); if (!sp) return NULL; - init_MUTEX(&sp->so_sema); - sp->so_seqid = 0; /* arbitrary */ INIT_LIST_HEAD(&sp->so_states); INIT_LIST_HEAD(&sp->so_delegations); + rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); + sp->so_seqid.sequence = &sp->so_sequence; + spin_lock_init(&sp->so_sequence.lock); + INIT_LIST_HEAD(&sp->so_sequence.list); atomic_set(&sp->so_count, 1); return sp; } @@ -359,7 +361,6 @@ nfs4_alloc_open_state(void) memset(state->stateid.data, 0, sizeof(state->stateid.data)); atomic_set(&state->count, 1); INIT_LIST_HEAD(&state->lock_states); - init_MUTEX(&state->lock_sema); spin_lock_init(&state->state_lock); return state; } @@ -441,7 +442,6 @@ nfs4_get_open_state(struct inode *inode, state = __nfs4_find_state_byowner(inode, owner); if (state == NULL && new != NULL) { state = new; - /* Caller *must* be holding owner->so_sem */ /* Note: The reclaim code dictates that we add stateless * and read-only stateids to the end of the list */ list_add_tail(&state->open_states, &owner->so_states); @@ -461,7 +461,7 @@ out: /* * Beware! Caller must be holding exactly one - * reference to clp->cl_sem and owner->so_sema! + * reference to clp->cl_sem! */ void nfs4_put_open_state(struct nfs4_state *state) { @@ -481,19 +481,15 @@ void nfs4_put_open_state(struct nfs4_sta } /* - * Beware! Caller must be holding no references to clp->cl_sem! - * of owner->so_sema! + * Close the current file. */ void nfs4_close_state(struct nfs4_state *state, mode_t mode) { struct inode *inode = state->inode; struct nfs4_state_owner *owner = state->owner; - struct nfs4_client *clp = owner->so_client; int newstate; atomic_inc(&owner->so_count); - down_read(&clp->cl_sem); - down(&owner->so_sema); /* Protect against nfs4_find_state() */ spin_lock(&inode->i_lock); if (mode & FMODE_READ) @@ -515,14 +511,16 @@ void nfs4_close_state(struct nfs4_state newstate |= FMODE_WRITE; if (state->state == newstate) goto out; - if (nfs4_do_close(inode, state, newstate) == -EINPROGRESS) + if (test_bit(NFS_DELEGATED_STATE, &state->flags)) { + state->state = newstate; + goto out; + } + if (nfs4_do_close(inode, state, newstate) == 0) return; } out: nfs4_put_open_state(state); - up(&owner->so_sema); nfs4_put_state_owner(owner); - up_read(&clp->cl_sem); } /* @@ -546,19 +544,16 @@ __nfs4_find_lock_state(struct nfs4_state * Return a compatible lock_state. If no initialized lock_state structure * exists, return an uninitialized one. * - * The caller must be holding state->lock_sema */ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) { struct nfs4_lock_state *lsp; struct nfs4_client *clp = state->owner->so_client; - lsp = kmalloc(sizeof(*lsp), GFP_KERNEL); + lsp = kzalloc(sizeof(*lsp), GFP_KERNEL); if (lsp == NULL) return NULL; - lsp->ls_flags = 0; - lsp->ls_seqid = 0; /* arbitrary */ - memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data)); + lsp->ls_seqid.sequence = &state->owner->so_sequence; atomic_set(&lsp->ls_count, 1); lsp->ls_owner = fl_owner; spin_lock(&clp->cl_lock); @@ -572,7 +567,7 @@ static struct nfs4_lock_state *nfs4_allo * Return a compatible lock_state. If no initialized lock_state structure * exists, return an uninitialized one. * - * The caller must be holding state->lock_sema and clp->cl_sem + * The caller must be holding clp->cl_sem */ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) { @@ -605,7 +600,7 @@ static struct nfs4_lock_state *nfs4_get_ * Release reference to lock_state, and free it if we see that * it is no longer in use */ -static void nfs4_put_lock_state(struct nfs4_lock_state *lsp) +void nfs4_put_lock_state(struct nfs4_lock_state *lsp) { struct nfs4_state *state; @@ -673,29 +668,98 @@ void nfs4_copy_stateid(nfs4_stateid *dst nfs4_put_lock_state(lsp); } -/* -* Called with state->lock_sema and clp->cl_sem held. -*/ -void nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp) -{ - if (status == NFS_OK || seqid_mutating_err(-status)) - lsp->ls_seqid++; +struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter) +{ + struct rpc_sequence *sequence = counter->sequence; + struct nfs_seqid *new; + + new = kmalloc(sizeof(*new), GFP_KERNEL); + if (new != NULL) { + new->sequence = counter; + new->task = NULL; + spin_lock(&sequence->lock); + list_add_tail(&new->list, &sequence->list); + spin_unlock(&sequence->lock); + } + return new; +} + +void nfs_free_seqid(struct nfs_seqid *seqid) +{ + struct rpc_sequence *sequence = seqid->sequence->sequence; + struct rpc_task *next = NULL; + + spin_lock(&sequence->lock); + list_del(&seqid->list); + if (!list_empty(&sequence->list)) { + next = list_entry(sequence->list.next, struct nfs_seqid, list)->task; + if (next) + rpc_wake_up_task(next); + } + spin_unlock(&sequence->lock); + kfree(seqid); } /* -* Called with sp->so_sema and clp->cl_sem held. -* -* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or -* failed with a seqid incrementing error - -* see comments nfs_fs.h:seqid_mutating_error() -*/ -void nfs4_increment_seqid(int status, struct nfs4_state_owner *sp) -{ - if (status == NFS_OK || seqid_mutating_err(-status)) - sp->so_seqid++; - /* If the server returns BAD_SEQID, unhash state_owner here */ - if (status == -NFS4ERR_BAD_SEQID) + * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or + * failed with a seqid incrementing error - + * see comments nfs_fs.h:seqid_mutating_error() + */ +static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid) +{ + switch (status) { + case 0: + break; + case -NFS4ERR_BAD_SEQID: + case -NFS4ERR_STALE_CLIENTID: + case -NFS4ERR_STALE_STATEID: + case -NFS4ERR_BAD_STATEID: + case -NFS4ERR_BADXDR: + case -NFS4ERR_RESOURCE: + case -NFS4ERR_NOFILEHANDLE: + /* Non-seqid mutating errors */ + return; + }; + /* + * Note: no locking needed as we are guaranteed to be first + * on the sequence list + */ + seqid->sequence->counter++; +} + +void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid) +{ + if (status == -NFS4ERR_BAD_SEQID) { + struct nfs4_state_owner *sp = container_of(seqid->sequence, + struct nfs4_state_owner, so_seqid); nfs4_drop_state_owner(sp); + } + return nfs_increment_seqid(status, seqid); +} + +/* + * Increment the seqid if the LOCK/LOCKU succeeded, or + * failed with a seqid incrementing error - + * see comments nfs_fs.h:seqid_mutating_error() + */ +void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid) +{ + return nfs_increment_seqid(status, seqid); +} + +int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task) +{ + struct rpc_sequence *sequence = seqid->sequence->sequence; + int status = 0; + + spin_lock(&sequence->lock); + if (sequence->list.next != &seqid->list) { + seqid->task = task; + rpc_sleep_on(&sequence->wait, task, NULL, NULL); + status = -EAGAIN; + } + spin_unlock(&sequence->lock); + return status; } static int reclaimer(void *); @@ -791,8 +855,6 @@ static int nfs4_reclaim_open_state(struc if (state->state == 0) continue; status = ops->recover_open(sp, state); - list_for_each_entry(lock, &state->lock_states, ls_locks) - lock->ls_flags &= ~NFS_LOCK_INITIALIZED; if (status >= 0) { status = nfs4_reclaim_locks(ops, state); if (status < 0) @@ -831,6 +893,26 @@ out_err: return status; } +static void nfs4_state_mark_reclaim(struct nfs4_client *clp) +{ + struct nfs4_state_owner *sp; + struct nfs4_state *state; + struct nfs4_lock_state *lock; + + /* Reset all sequence ids to zero */ + list_for_each_entry(sp, &clp->cl_state_owners, so_list) { + sp->so_seqid.counter = 0; + sp->so_seqid.flags = 0; + list_for_each_entry(state, &sp->so_states, open_states) { + list_for_each_entry(lock, &state->lock_states, ls_locks) { + lock->ls_seqid.counter = 0; + lock->ls_seqid.flags = 0; + lock->ls_flags &= ~NFS_LOCK_INITIALIZED; + } + } + } +} + static int reclaimer(void *ptr) { struct reclaimer_args *args = (struct reclaimer_args *)ptr; @@ -864,6 +946,7 @@ restart_loop: default: ops = &nfs4_network_partition_recovery_ops; }; + nfs4_state_mark_reclaim(clp); status = __nfs4_init_client(clp); if (status) goto out_error; Index: linux-2.6/fs/nfs/nfs4xdr.c =================================================================== --- linux-2.6.orig/fs/nfs/nfs4xdr.c +++ linux-2.6/fs/nfs/nfs4xdr.c @@ -602,10 +602,10 @@ static int encode_close(struct xdr_strea { uint32_t *p; - RESERVE_SPACE(8+sizeof(arg->stateid.data)); + RESERVE_SPACE(8+sizeof(arg->stateid->data)); WRITE32(OP_CLOSE); - WRITE32(arg->seqid); - WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); + WRITE32(arg->seqid->sequence->counter); + WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data)); return 0; } @@ -729,22 +729,18 @@ static int encode_lock(struct xdr_stream WRITE64(arg->length); WRITE32(opargs->new_lock_owner); if (opargs->new_lock_owner){ - struct nfs_open_to_lock *ol = opargs->u.open_lock; - RESERVE_SPACE(40); - WRITE32(ol->open_seqid); - WRITEMEM(&ol->open_stateid, sizeof(ol->open_stateid)); - WRITE32(ol->lock_seqid); - WRITE64(ol->lock_owner.clientid); + WRITE32(opargs->open_seqid->sequence->counter); + WRITEMEM(opargs->open_stateid->data, sizeof(opargs->open_stateid->data)); + WRITE32(opargs->lock_seqid->sequence->counter); + WRITE64(opargs->lock_owner.clientid); WRITE32(4); - WRITE32(ol->lock_owner.id); + WRITE32(opargs->lock_owner.id); } else { - struct nfs_exist_lock *el = opargs->u.exist_lock; - RESERVE_SPACE(20); - WRITEMEM(&el->stateid, sizeof(el->stateid)); - WRITE32(el->seqid); + WRITEMEM(opargs->lock_stateid->data, sizeof(opargs->lock_stateid->data)); + WRITE32(opargs->lock_seqid->sequence->counter); } return 0; @@ -775,8 +771,8 @@ static int encode_locku(struct xdr_strea RESERVE_SPACE(44); WRITE32(OP_LOCKU); WRITE32(arg->type); - WRITE32(opargs->seqid); - WRITEMEM(&opargs->stateid, sizeof(opargs->stateid)); + WRITE32(opargs->seqid->sequence->counter); + WRITEMEM(opargs->stateid->data, sizeof(opargs->stateid->data)); WRITE64(arg->offset); WRITE64(arg->length); @@ -826,7 +822,7 @@ static inline void encode_openhdr(struct */ RESERVE_SPACE(8); WRITE32(OP_OPEN); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); RESERVE_SPACE(16); WRITE64(arg->clientid); @@ -941,7 +937,7 @@ static int encode_open_confirm(struct xd RESERVE_SPACE(8+sizeof(arg->stateid.data)); WRITE32(OP_OPEN_CONFIRM); WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); - WRITE32(arg->seqid); + WRITE32(arg->seqid->sequence->counter); return 0; } @@ -950,10 +946,10 @@ static int encode_open_downgrade(struct { uint32_t *p; - RESERVE_SPACE(8+sizeof(arg->stateid.data)); + RESERVE_SPACE(8+sizeof(arg->stateid->data)); WRITE32(OP_OPEN_DOWNGRADE); - WRITEMEM(arg->stateid.data, sizeof(arg->stateid.data)); - WRITE32(arg->seqid); + WRITEMEM(arg->stateid->data, sizeof(arg->stateid->data)); + WRITE32(arg->seqid->sequence->counter); encode_share_access(xdr, arg->open_flags); return 0; } @@ -1437,6 +1433,9 @@ static int nfs4_xdr_enc_open(struct rpc_ }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1464,6 +1463,9 @@ static int nfs4_xdr_enc_open_confirm(str }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1485,6 +1487,9 @@ static int nfs4_xdr_enc_open_noattr(stru }; int status; + status = nfs_wait_on_sequence(args->seqid, req->rq_task); + if (status != 0) + goto out; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -1525,8 +1530,15 @@ static int nfs4_xdr_enc_lock(struct rpc_ struct compound_hdr hdr = { .nops = 2, }; + struct nfs_lock_opargs *opargs = args->u.lock; int status; + status = nfs_wait_on_sequence(opargs->lock_seqid, req->rq_task); + if (status != 0) + goto out; + /* Do we need to do an open_to_lock_owner? */ + if (opargs->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED) + opargs->new_lock_owner = 0; xdr_init_encode(&xdr, &req->rq_snd_buf, p); encode_compound_hdr(&xdr, &hdr); status = encode_putfh(&xdr, args->fh); @@ -2890,8 +2902,8 @@ static int decode_lock(struct xdr_stream status = decode_op_hdr(xdr, OP_LOCK); if (status == 0) { - READ_BUF(sizeof(nfs4_stateid)); - COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); + READ_BUF(sizeof(res->u.stateid.data)); + COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); } else if (status == -NFS4ERR_DENIED) return decode_lock_denied(xdr, &res->u.denied); return status; @@ -2913,8 +2925,8 @@ static int decode_locku(struct xdr_strea status = decode_op_hdr(xdr, OP_LOCKU); if (status == 0) { - READ_BUF(sizeof(nfs4_stateid)); - COPYMEM(&res->u.stateid, sizeof(res->u.stateid)); + READ_BUF(sizeof(res->u.stateid.data)); + COPYMEM(res->u.stateid.data, sizeof(res->u.stateid.data)); } return status; } Index: linux-2.6/fs/nfs/proc.c =================================================================== --- linux-2.6.orig/fs/nfs/proc.c +++ linux-2.6/fs/nfs/proc.c @@ -216,7 +216,7 @@ static int nfs_proc_write(struct nfs_wri static int nfs_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr, - int flags) + int flags, struct nameidata *nd) { struct nfs_fh fhandle; struct nfs_fattr fattr; Index: linux-2.6/fs/open.c =================================================================== --- linux-2.6.orig/fs/open.c +++ linux-2.6/fs/open.c @@ -739,7 +739,8 @@ asmlinkage long sys_fchown(unsigned int } static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, - int flags, struct file *f) + int flags, struct file *f, + int (*open)(struct inode *, struct file *)) { struct inode *inode; int error; @@ -761,11 +762,14 @@ static struct file *__dentry_open(struct f->f_op = fops_get(inode->i_fop); file_move(f, &inode->i_sb->s_files); - if (f->f_op && f->f_op->open) { - error = f->f_op->open(inode,f); + if (!open && f->f_op) + open = f->f_op->open; + if (open) { + error = open(inode, f); if (error) goto cleanup_all; } + f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); @@ -814,28 +818,75 @@ struct file *filp_open(const char * file { int namei_flags, error; struct nameidata nd; - struct file *f; namei_flags = flags; if ((namei_flags+1) & O_ACCMODE) namei_flags++; - if (namei_flags & O_TRUNC) - namei_flags |= 2; - - error = -ENFILE; - f = get_empty_filp(); - if (f == NULL) - return ERR_PTR(error); error = open_namei(filename, namei_flags, mode, &nd); if (!error) - return __dentry_open(nd.dentry, nd.mnt, flags, f); + return nameidata_to_filp(&nd, flags); - put_filp(f); return ERR_PTR(error); } EXPORT_SYMBOL(filp_open); +/** + * lookup_instantiate_filp - instantiates the open intent filp + * @nd: pointer to nameidata + * @dentry: pointer to dentry + * @open: open callback + * + * Helper for filesystems that want to use lookup open intents and pass back + * a fully instantiated struct file to the caller. + * This function is meant to be called from within a filesystem's + * lookup method. + * Note that in case of error, nd->intent.open.file is destroyed, but the + * path information remains valid. + * If the open callback is set to NULL, then the standard f_op->open() + * filesystem callback is substituted. + */ +struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry, + int (*open)(struct inode *, struct file *)) +{ + if (IS_ERR(nd->intent.open.file)) + goto out; + if (IS_ERR(dentry)) + goto out_err; + nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt), + nd->intent.open.flags - 1, + nd->intent.open.file, + open); +out: + return nd->intent.open.file; +out_err: + release_open_intent(nd); + nd->intent.open.file = (struct file *)dentry; + goto out; +} +EXPORT_SYMBOL_GPL(lookup_instantiate_filp); + +/** + * nameidata_to_filp - convert a nameidata to an open filp. + * @nd: pointer to nameidata + * @flags: open flags + * + * Note that this function destroys the original nameidata + */ +struct file *nameidata_to_filp(struct nameidata *nd, int flags) +{ + struct file *filp; + + /* Pick up the filp from the open intent */ + filp = nd->intent.open.file; + /* Has the filesystem initialised the file for us? */ + if (filp->f_dentry == NULL) + filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL); + else + path_release(nd); + return filp; +} + struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) { int error; @@ -846,7 +897,7 @@ struct file *dentry_open(struct dentry * if (f == NULL) return ERR_PTR(error); - return __dentry_open(dentry, mnt, flags, f); + return __dentry_open(dentry, mnt, flags, f, NULL); } EXPORT_SYMBOL(dentry_open); Index: linux-2.6/include/linux/namei.h =================================================================== --- linux-2.6.orig/include/linux/namei.h +++ linux-2.6/include/linux/namei.h @@ -8,6 +8,7 @@ struct vfsmount; struct open_intent { int flags; int create_mode; + struct file *file; }; enum { MAX_NESTED_LINKS = 5 }; @@ -65,6 +66,13 @@ extern int FASTCALL(link_path_walk(const extern void path_release(struct nameidata *); extern void path_release_on_umount(struct nameidata *); +extern int __user_path_lookup_open(const char __user *, unsigned lookup_flags, struct nameidata *nd, int open_flags); +extern int path_lookup_open(const char *, unsigned lookup_flags, struct nameidata *, int open_flags); +extern struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry, + int (*open)(struct inode *, struct file *)); +extern struct file *nameidata_to_filp(struct nameidata *nd, int flags); +extern void release_open_intent(struct nameidata *); + extern struct dentry * lookup_one_len(const char *, struct dentry *, int); extern struct dentry * lookup_hash(struct qstr *, struct dentry *); Index: linux-2.6/include/linux/nfs_fs.h =================================================================== --- linux-2.6.orig/include/linux/nfs_fs.h +++ linux-2.6/include/linux/nfs_fs.h @@ -41,6 +41,10 @@ #define NFS_MAX_FILE_IO_BUFFER_SIZE 32768 #define NFS_DEF_FILE_IO_BUFFER_SIZE 4096 +/* Default timeout values */ +#define NFS_MAX_UDP_TIMEOUT (60*HZ) +#define NFS_MAX_TCP_TIMEOUT (600*HZ) + /* * superblock magic number for NFS */ Index: linux-2.6/include/linux/nfs_xdr.h =================================================================== --- linux-2.6.orig/include/linux/nfs_xdr.h +++ linux-2.6/include/linux/nfs_xdr.h @@ -96,12 +96,13 @@ struct nfs4_change_info { u64 after; }; +struct nfs_seqid; /* * Arguments to the open call. */ struct nfs_openargs { const struct nfs_fh * fh; - __u32 seqid; + struct nfs_seqid * seqid; int open_flags; __u64 clientid; __u32 id; @@ -136,7 +137,7 @@ struct nfs_openres { struct nfs_open_confirmargs { const struct nfs_fh * fh; nfs4_stateid stateid; - __u32 seqid; + struct nfs_seqid * seqid; }; struct nfs_open_confirmres { @@ -148,8 +149,8 @@ struct nfs_open_confirmres { */ struct nfs_closeargs { struct nfs_fh * fh; - nfs4_stateid stateid; - __u32 seqid; + nfs4_stateid * stateid; + struct nfs_seqid * seqid; int open_flags; }; @@ -164,30 +165,19 @@ struct nfs_lowner { u32 id; }; -struct nfs_open_to_lock { - __u32 open_seqid; - nfs4_stateid open_stateid; - __u32 lock_seqid; - struct nfs_lowner lock_owner; -}; - -struct nfs_exist_lock { - nfs4_stateid stateid; - __u32 seqid; -}; - struct nfs_lock_opargs { + struct nfs_seqid * lock_seqid; + nfs4_stateid * lock_stateid; + struct nfs_seqid * open_seqid; + nfs4_stateid * open_stateid; + struct nfs_lowner lock_owner; __u32 reclaim; __u32 new_lock_owner; - union { - struct nfs_open_to_lock *open_lock; - struct nfs_exist_lock *exist_lock; - } u; }; struct nfs_locku_opargs { - __u32 seqid; - nfs4_stateid stateid; + struct nfs_seqid * seqid; + nfs4_stateid * stateid; }; struct nfs_lockargs { @@ -722,7 +712,7 @@ struct nfs_rpc_ops { int (*write) (struct nfs_write_data *); int (*commit) (struct nfs_write_data *); int (*create) (struct inode *, struct dentry *, - struct iattr *, int); + struct iattr *, int, struct nameidata *); int (*remove) (struct inode *, struct qstr *); int (*unlink_setup) (struct rpc_message *, struct dentry *, struct qstr *); Index: linux-2.6/include/linux/sunrpc/debug.h =================================================================== --- linux-2.6.orig/include/linux/sunrpc/debug.h +++ linux-2.6/include/linux/sunrpc/debug.h @@ -32,6 +32,7 @@ #define RPCDBG_AUTH 0x0010 #define RPCDBG_PMAP 0x0020 #define RPCDBG_SCHED 0x0040 +#define RPCDBG_TRANS 0x0080 #define RPCDBG_SVCSOCK 0x0100 #define RPCDBG_SVCDSP 0x0200 #define RPCDBG_MISC 0x0400 @@ -94,6 +95,8 @@ enum { CTL_NLMDEBUG, CTL_SLOTTABLE_UDP, CTL_SLOTTABLE_TCP, + CTL_MIN_RESVPORT, + CTL_MAX_RESVPORT, }; #endif /* _LINUX_SUNRPC_DEBUG_H_ */ Index: linux-2.6/include/linux/sunrpc/msg_prot.h =================================================================== --- linux-2.6.orig/include/linux/sunrpc/msg_prot.h +++ linux-2.6/include/linux/sunrpc/msg_prot.h @@ -76,5 +76,30 @@ enum rpc_auth_stat { #define RPC_MAXNETNAMELEN 256 +/* + * From RFC 1831: + * + * "A record is composed of one or more record fragments. A record + * fragment is a four-byte header followed by 0 to (2**31) - 1 bytes of + * fragment data. The bytes encode an unsigned binary number; as with + * XDR integers, the byte order is from highest to lowest. The number + * encodes two values -- a boolean which indicates whether the fragment + * is the last fragment of the record (bit value 1 implies the fragment + * is the last fragment) and a 31-bit unsigned binary value which is the + * length in bytes of the fragment's data. The boolean value is the + * highest-order bit of the header; the length is the 31 low-order bits. + * (Note that this record specification is NOT in XDR standard form!)" + * + * The Linux RPC client always sends its requests in a single record + * fragment, limiting the maximum payload size for stream transports to + * 2GB. + */ + +typedef u32 rpc_fraghdr; + +#define RPC_LAST_STREAM_FRAGMENT (1U << 31) +#define RPC_FRAGMENT_SIZE_MASK (~RPC_LAST_STREAM_FRAGMENT) +#define RPC_MAX_FRAGMENT_SIZE ((1U << 31) - 1) + #endif /* __KERNEL__ */ #endif /* _LINUX_SUNRPC_MSGPROT_H_ */ Index: linux-2.6/include/linux/sunrpc/xdr.h =================================================================== --- linux-2.6.orig/include/linux/sunrpc/xdr.h +++ linux-2.6/include/linux/sunrpc/xdr.h @@ -161,14 +161,10 @@ typedef struct { typedef size_t (*skb_read_actor_t)(skb_reader_t *desc, void *to, size_t len); +extern int csum_partial_copy_to_xdr(struct xdr_buf *, struct sk_buff *); extern ssize_t xdr_partial_copy_from_skb(struct xdr_buf *, unsigned int, skb_reader_t *, skb_read_actor_t); -struct socket; -struct sockaddr; -extern int xdr_sendpages(struct socket *, struct sockaddr *, int, - struct xdr_buf *, unsigned int, int); - extern int xdr_encode_word(struct xdr_buf *, int, u32); extern int xdr_decode_word(struct xdr_buf *, int, u32 *); Index: linux-2.6/include/linux/sunrpc/xprt.h =================================================================== --- linux-2.6.orig/include/linux/sunrpc/xprt.h +++ linux-2.6/include/linux/sunrpc/xprt.h @@ -1,5 +1,5 @@ /* - * linux/include/linux/sunrpc/clnt_xprt.h + * linux/include/linux/sunrpc/xprt.h * * Declarations for the RPC transport interface. * @@ -15,20 +15,6 @@ #include #include -/* - * The transport code maintains an estimate on the maximum number of out- - * standing RPC requests, using a smoothed version of the congestion - * avoidance implemented in 44BSD. This is basically the Van Jacobson - * congestion algorithm: If a retransmit occurs, the congestion window is - * halved; otherwise, it is incremented by 1/cwnd when - * - * - a reply is received and - * - a full number of requests are outstanding and - * - the congestion window hasn't been updated recently. - * - * Upper procedures may check whether a request would block waiting for - * a free RPC slot by using the RPC_CONGESTED() macro. - */ extern unsigned int xprt_udp_slot_table_entries; extern unsigned int xprt_tcp_slot_table_entries; @@ -36,34 +22,23 @@ extern unsigned int xprt_tcp_slot_table_ #define RPC_DEF_SLOT_TABLE (16U) #define RPC_MAX_SLOT_TABLE (128U) -#define RPC_CWNDSHIFT (8U) -#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT) -#define RPC_INITCWND RPC_CWNDSCALE -#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT) -#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd) - -/* Default timeout values */ -#define RPC_MAX_UDP_TIMEOUT (60*HZ) -#define RPC_MAX_TCP_TIMEOUT (600*HZ) - /* - * Wait duration for an RPC TCP connection to be established. Solaris - * NFS over TCP uses 60 seconds, for example, which is in line with how - * long a server takes to reboot. + * RPC call and reply header size as number of 32bit words (verifier + * size computed separately) */ -#define RPC_CONNECT_TIMEOUT (60*HZ) +#define RPC_CALLHDRSIZE 6 +#define RPC_REPHDRSIZE 4 /* - * Delay an arbitrary number of seconds before attempting to reconnect - * after an error. + * Parameters for choosing a free port */ -#define RPC_REESTABLISH_TIMEOUT (15*HZ) +extern unsigned int xprt_min_resvport; +extern unsigned int xprt_max_resvport; -/* RPC call and reply header size as number of 32bit words (verifier - * size computed separately) - */ -#define RPC_CALLHDRSIZE 6 -#define RPC_REPHDRSIZE 4 +#define RPC_MIN_RESVPORT (1U) +#define RPC_MAX_RESVPORT (65535U) +#define RPC_DEF_MIN_RESVPORT (650U) +#define RPC_DEF_MAX_RESVPORT (1023U) /* * This describes a timeout strategy @@ -76,6 +51,9 @@ struct rpc_timeout { unsigned char to_exponential; }; +struct rpc_task; +struct rpc_xprt; + /* * This describes a complete RPC request */ @@ -121,12 +99,21 @@ struct rpc_rqst { #define rq_svec rq_snd_buf.head #define rq_slen rq_snd_buf.len -#define XPRT_LAST_FRAG (1 << 0) -#define XPRT_COPY_RECM (1 << 1) -#define XPRT_COPY_XID (1 << 2) -#define XPRT_COPY_DATA (1 << 3) +struct rpc_xprt_ops { + void (*set_buffer_size)(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize); + int (*reserve_xprt)(struct rpc_task *task); + void (*release_xprt)(struct rpc_xprt *xprt, struct rpc_task *task); + void (*connect)(struct rpc_task *task); + int (*send_request)(struct rpc_task *task); + void (*set_retrans_timeout)(struct rpc_task *task); + void (*timer)(struct rpc_task *task); + void (*release_request)(struct rpc_task *task); + void (*close)(struct rpc_xprt *xprt); + void (*destroy)(struct rpc_xprt *xprt); +}; struct rpc_xprt { + struct rpc_xprt_ops * ops; /* transport methods */ struct socket * sock; /* BSD socket layer */ struct sock * inet; /* INET layer */ @@ -137,11 +124,13 @@ struct rpc_xprt { unsigned long cong; /* current congestion */ unsigned long cwnd; /* congestion window */ - unsigned int rcvsize, /* socket receive buffer size */ - sndsize; /* socket send buffer size */ + size_t rcvsize, /* transport rcv buffer size */ + sndsize; /* transport send buffer size */ size_t max_payload; /* largest RPC payload size, in bytes */ + unsigned int tsh_size; /* size of transport specific + header */ struct rpc_wait_queue sending; /* requests waiting to send */ struct rpc_wait_queue resend; /* requests waiting to resend */ @@ -150,11 +139,9 @@ struct rpc_xprt { struct list_head free; /* free slots */ struct rpc_rqst * slot; /* slot table storage */ unsigned int max_reqs; /* total slots */ - unsigned long sockstate; /* Socket state */ + unsigned long state; /* transport state */ unsigned char shutdown : 1, /* being shut down */ - nocong : 1, /* no congestion control */ - resvport : 1, /* use a reserved port */ - stream : 1; /* TCP */ + resvport : 1; /* use a reserved port */ /* * XID @@ -171,22 +158,27 @@ struct rpc_xprt { unsigned long tcp_copied, /* copied to request */ tcp_flags; /* - * Connection of sockets + * Connection of transports */ - struct work_struct sock_connect; + unsigned long connect_timeout, + bind_timeout, + reestablish_timeout; + struct work_struct connect_worker; unsigned short port; + /* - * Disconnection of idle sockets + * Disconnection of idle transports */ struct work_struct task_cleanup; struct timer_list timer; - unsigned long last_used; + unsigned long last_used, + idle_timeout; /* * Send stuff */ - spinlock_t sock_lock; /* lock socket info */ - spinlock_t xprt_lock; /* lock xprt info */ + spinlock_t transport_lock; /* lock transport info */ + spinlock_t reserve_lock; /* lock slot table */ struct rpc_task * snd_task; /* Task blocked in send */ struct list_head recv; @@ -195,37 +187,111 @@ struct rpc_xprt { void (*old_data_ready)(struct sock *, int); void (*old_state_change)(struct sock *); void (*old_write_space)(struct sock *); - - wait_queue_head_t cong_wait; }; +#define XPRT_LAST_FRAG (1 << 0) +#define XPRT_COPY_RECM (1 << 1) +#define XPRT_COPY_XID (1 << 2) +#define XPRT_COPY_DATA (1 << 3) + #ifdef __KERNEL__ -struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr, - struct rpc_timeout *toparms); -int xprt_destroy(struct rpc_xprt *); -void xprt_set_timeout(struct rpc_timeout *, unsigned int, - unsigned long); - -void xprt_reserve(struct rpc_task *); -int xprt_prepare_transmit(struct rpc_task *); -void xprt_transmit(struct rpc_task *); -void xprt_receive(struct rpc_task *); +/* + * Transport operations used by ULPs + */ +struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr, struct rpc_timeout *to); +void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr); + +/* + * Generic internal transport functions + */ +void xprt_connect(struct rpc_task *task); +void xprt_reserve(struct rpc_task *task); +int xprt_reserve_xprt(struct rpc_task *task); +int xprt_reserve_xprt_cong(struct rpc_task *task); +int xprt_prepare_transmit(struct rpc_task *task); +void xprt_transmit(struct rpc_task *task); +void xprt_abort_transmit(struct rpc_task *task); int xprt_adjust_timeout(struct rpc_rqst *req); -void xprt_release(struct rpc_task *); -void xprt_connect(struct rpc_task *); -void xprt_sock_setbufsize(struct rpc_xprt *); - -#define XPRT_LOCKED 0 -#define XPRT_CONNECT 1 -#define XPRT_CONNECTING 2 - -#define xprt_connected(xp) (test_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_set_connected(xp) (set_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_test_and_set_connected(xp) (test_and_set_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_test_and_clear_connected(xp) \ - (test_and_clear_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_clear_connected(xp) (clear_bit(XPRT_CONNECT, &(xp)->sockstate)) +void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task); +void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task); +void xprt_release(struct rpc_task *task); +int xprt_destroy(struct rpc_xprt *xprt); + +static inline u32 *xprt_skip_transport_header(struct rpc_xprt *xprt, u32 *p) +{ + return p + xprt->tsh_size; +} + +/* + * Transport switch helper functions + */ +void xprt_set_retrans_timeout_def(struct rpc_task *task); +void xprt_set_retrans_timeout_rtt(struct rpc_task *task); +void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status); +void xprt_wait_for_buffer_space(struct rpc_task *task); +void xprt_write_space(struct rpc_xprt *xprt); +void xprt_update_rtt(struct rpc_task *task); +void xprt_adjust_cwnd(struct rpc_task *task, int result); +struct rpc_rqst * xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid); +void xprt_complete_rqst(struct rpc_task *task, int copied); +void xprt_release_rqst_cong(struct rpc_task *task); +void xprt_disconnect(struct rpc_xprt *xprt); + +/* + * Socket transport setup operations + */ +int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to); +int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to); + +/* + * Reserved bit positions in xprt->state + */ +#define XPRT_LOCKED (0) +#define XPRT_CONNECTED (1) +#define XPRT_CONNECTING (2) + +static inline void xprt_set_connected(struct rpc_xprt *xprt) +{ + set_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline void xprt_clear_connected(struct rpc_xprt *xprt) +{ + clear_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_connected(struct rpc_xprt *xprt) +{ + return test_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_test_and_set_connected(struct rpc_xprt *xprt) +{ + return test_and_set_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_test_and_clear_connected(struct rpc_xprt *xprt) +{ + return test_and_clear_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline void xprt_clear_connecting(struct rpc_xprt *xprt) +{ + smp_mb__before_clear_bit(); + clear_bit(XPRT_CONNECTING, &xprt->state); + smp_mb__after_clear_bit(); +} + +static inline int xprt_connecting(struct rpc_xprt *xprt) +{ + return test_bit(XPRT_CONNECTING, &xprt->state); +} + +static inline int xprt_test_and_set_connecting(struct rpc_xprt *xprt) +{ + return test_and_set_bit(XPRT_CONNECTING, &xprt->state); +} #endif /* __KERNEL__*/ Index: linux-2.6/net/sunrpc/Makefile =================================================================== --- linux-2.6.orig/net/sunrpc/Makefile +++ linux-2.6/net/sunrpc/Makefile @@ -6,7 +6,7 @@ obj-$(CONFIG_SUNRPC) += sunrpc.o obj-$(CONFIG_SUNRPC_GSS) += auth_gss/ -sunrpc-y := clnt.o xprt.o sched.o \ +sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \ auth.o auth_null.o auth_unix.o \ svc.o svcsock.o svcauth.o svcauth_unix.o \ pmap_clnt.o timer.o xdr.o \ Index: linux-2.6/net/sunrpc/auth.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth.c +++ linux-2.6/net/sunrpc/auth.c @@ -11,7 +11,6 @@ #include #include #include -#include #include #include Index: linux-2.6/net/sunrpc/auth_gss/auth_gss.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth_gss/auth_gss.c +++ linux-2.6/net/sunrpc/auth_gss/auth_gss.c @@ -42,8 +42,6 @@ #include #include #include -#include -#include #include #include #include @@ -846,10 +844,8 @@ gss_marshal(struct rpc_task *task, u32 * /* We compute the checksum for the verifier over the xdr-encoded bytes * starting with the xid and ending at the end of the credential: */ - iov.iov_base = req->rq_snd_buf.head[0].iov_base; - if (task->tk_client->cl_xprt->stream) - /* See clnt.c:call_header() */ - iov.iov_base += 4; + iov.iov_base = xprt_skip_transport_header(task->tk_xprt, + req->rq_snd_buf.head[0].iov_base); iov.iov_len = (u8 *)p - (u8 *)iov.iov_base; xdr_buf_from_iov(&iov, &verf_buf); Index: linux-2.6/net/sunrpc/auth_gss/gss_krb5_mech.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth_gss/gss_krb5_mech.c +++ linux-2.6/net/sunrpc/auth_gss/gss_krb5_mech.c @@ -39,7 +39,6 @@ #include #include #include -#include #include #include #include Index: linux-2.6/net/sunrpc/auth_gss/gss_mech_switch.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth_gss/gss_mech_switch.c +++ linux-2.6/net/sunrpc/auth_gss/gss_mech_switch.c @@ -35,7 +35,6 @@ #include #include -#include #include #include #include Index: linux-2.6/net/sunrpc/auth_null.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth_null.c +++ linux-2.6/net/sunrpc/auth_null.c @@ -7,9 +7,7 @@ */ #include -#include #include -#include #include #include #include Index: linux-2.6/net/sunrpc/auth_unix.c =================================================================== --- linux-2.6.orig/net/sunrpc/auth_unix.c +++ linux-2.6/net/sunrpc/auth_unix.c @@ -9,8 +9,6 @@ #include #include #include -#include -#include #include #include Index: linux-2.6/net/sunrpc/clnt.c =================================================================== --- linux-2.6.orig/net/sunrpc/clnt.c +++ linux-2.6/net/sunrpc/clnt.c @@ -1,5 +1,5 @@ /* - * linux/net/sunrpc/rpcclnt.c + * linux/net/sunrpc/clnt.c * * This file contains the high-level RPC interface. * It is modeled as a finite state machine to support both synchronous @@ -27,7 +27,6 @@ #include #include #include -#include #include #include @@ -53,6 +52,7 @@ static void call_allocate(struct rpc_tas static void call_encode(struct rpc_task *task); static void call_decode(struct rpc_task *task); static void call_bind(struct rpc_task *task); +static void call_bind_status(struct rpc_task *task); static void call_transmit(struct rpc_task *task); static void call_status(struct rpc_task *task); static void call_refresh(struct rpc_task *task); @@ -517,15 +517,8 @@ void rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize) { struct rpc_xprt *xprt = clnt->cl_xprt; - - xprt->sndsize = 0; - if (sndsize) - xprt->sndsize = sndsize + RPC_SLACK_SPACE; - xprt->rcvsize = 0; - if (rcvsize) - xprt->rcvsize = rcvsize + RPC_SLACK_SPACE; - if (xprt_connected(xprt)) - xprt_sock_setbufsize(xprt); + if (xprt->ops->set_buffer_size) + xprt->ops->set_buffer_size(xprt, sndsize, rcvsize); } /* @@ -685,13 +678,11 @@ call_allocate(struct rpc_task *task) static void call_encode(struct rpc_task *task) { - struct rpc_clnt *clnt = task->tk_client; struct rpc_rqst *req = task->tk_rqstp; struct xdr_buf *sndbuf = &req->rq_snd_buf; struct xdr_buf *rcvbuf = &req->rq_rcv_buf; unsigned int bufsiz; kxdrproc_t encode; - int status; u32 *p; dprintk("RPC: %4d call_encode (status %d)\n", @@ -719,12 +710,9 @@ call_encode(struct rpc_task *task) rpc_exit(task, -EIO); return; } - if (encode && (status = rpcauth_wrap_req(task, encode, req, p, - task->tk_msg.rpc_argp)) < 0) { - printk(KERN_WARNING "%s: can't encode arguments: %d\n", - clnt->cl_protname, -status); - rpc_exit(task, status); - } + if (encode != NULL) + task->tk_status = rpcauth_wrap_req(task, encode, req, p, + task->tk_msg.rpc_argp); } /* @@ -734,43 +722,95 @@ static void call_bind(struct rpc_task *task) { struct rpc_clnt *clnt = task->tk_client; - struct rpc_xprt *xprt = clnt->cl_xprt; - - dprintk("RPC: %4d call_bind xprt %p %s connected\n", task->tk_pid, - xprt, (xprt_connected(xprt) ? "is" : "is not")); - task->tk_action = (xprt_connected(xprt)) ? call_transmit : call_connect; + dprintk("RPC: %4d call_bind (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_action = call_connect; if (!clnt->cl_port) { - task->tk_action = call_connect; - task->tk_timeout = RPC_CONNECT_TIMEOUT; + task->tk_action = call_bind_status; + task->tk_timeout = task->tk_xprt->bind_timeout; rpc_getport(task, clnt); } } /* - * 4a. Connect to the RPC server (TCP case) + * 4a. Sort out bind result */ static void -call_connect(struct rpc_task *task) +call_bind_status(struct rpc_task *task) { - struct rpc_clnt *clnt = task->tk_client; + int status = -EACCES; - dprintk("RPC: %4d call_connect status %d\n", - task->tk_pid, task->tk_status); - - if (xprt_connected(clnt->cl_xprt)) { - task->tk_action = call_transmit; + if (task->tk_status >= 0) { + dprintk("RPC: %4d call_bind_status (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_status = 0; + task->tk_action = call_connect; return; } - task->tk_action = call_connect_status; - if (task->tk_status < 0) - return; - xprt_connect(task); + + switch (task->tk_status) { + case -EACCES: + dprintk("RPC: %4d remote rpcbind: RPC program/version unavailable\n", + task->tk_pid); + rpc_delay(task, 3*HZ); + goto retry_bind; + case -ETIMEDOUT: + dprintk("RPC: %4d rpcbind request timed out\n", + task->tk_pid); + if (RPC_IS_SOFT(task)) { + status = -EIO; + break; + } + goto retry_bind; + case -EPFNOSUPPORT: + dprintk("RPC: %4d remote rpcbind service unavailable\n", + task->tk_pid); + break; + case -EPROTONOSUPPORT: + dprintk("RPC: %4d remote rpcbind version 2 unavailable\n", + task->tk_pid); + break; + default: + dprintk("RPC: %4d unrecognized rpcbind error (%d)\n", + task->tk_pid, -task->tk_status); + status = -EIO; + break; + } + + rpc_exit(task, status); + return; + +retry_bind: + task->tk_status = 0; + task->tk_action = call_bind; + return; } /* - * 4b. Sort out connect result + * 4b. Connect to the RPC server + */ +static void +call_connect(struct rpc_task *task) +{ + struct rpc_xprt *xprt = task->tk_xprt; + + dprintk("RPC: %4d call_connect xprt %p %s connected\n", + task->tk_pid, xprt, + (xprt_connected(xprt) ? "is" : "is not")); + + task->tk_action = call_transmit; + if (!xprt_connected(xprt)) { + task->tk_action = call_connect_status; + if (task->tk_status < 0) + return; + xprt_connect(task); + } +} + +/* + * 4c. Sort out connect result */ static void call_connect_status(struct rpc_task *task) @@ -778,6 +818,9 @@ call_connect_status(struct rpc_task *tas struct rpc_clnt *clnt = task->tk_client; int status = task->tk_status; + dprintk("RPC: %5u call_connect_status (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_status = 0; if (status >= 0) { clnt->cl_stats->netreconn++; @@ -785,17 +828,19 @@ call_connect_status(struct rpc_task *tas return; } - /* Something failed: we may have to rebind */ + /* Something failed: remote service port may have changed */ if (clnt->cl_autobind) clnt->cl_port = 0; + switch (status) { case -ENOTCONN: case -ETIMEDOUT: case -EAGAIN: - task->tk_action = (clnt->cl_port == 0) ? call_bind : call_connect; + task->tk_action = call_bind; break; default: rpc_exit(task, -EIO); + break; } } @@ -815,10 +860,12 @@ call_transmit(struct rpc_task *task) if (task->tk_status != 0) return; /* Encode here so that rpcsec_gss can use correct sequence number. */ - if (!task->tk_rqstp->rq_bytes_sent) + if (task->tk_rqstp->rq_bytes_sent == 0) { call_encode(task); - if (task->tk_status < 0) - return; + /* Did the encode result in an error condition? */ + if (task->tk_status != 0) + goto out_nosend; + } xprt_transmit(task); if (task->tk_status < 0) return; @@ -826,6 +873,10 @@ call_transmit(struct rpc_task *task) task->tk_action = NULL; rpc_wake_up_task(task); } + return; +out_nosend: + /* release socket write lock before attempting to handle error */ + xprt_abort_transmit(task); } /* @@ -1020,13 +1071,12 @@ static u32 * call_header(struct rpc_task *task) { struct rpc_clnt *clnt = task->tk_client; - struct rpc_xprt *xprt = clnt->cl_xprt; struct rpc_rqst *req = task->tk_rqstp; u32 *p = req->rq_svec[0].iov_base; /* FIXME: check buffer size? */ - if (xprt->stream) - *p++ = 0; /* fill in later */ + + p = xprt_skip_transport_header(task->tk_xprt, p); *p++ = req->rq_xid; /* XID */ *p++ = htonl(RPC_CALL); /* CALL */ *p++ = htonl(RPC_VERSION); /* RPC version */ Index: linux-2.6/net/sunrpc/pmap_clnt.c =================================================================== --- linux-2.6.orig/net/sunrpc/pmap_clnt.c +++ linux-2.6/net/sunrpc/pmap_clnt.c @@ -26,7 +26,7 @@ #define PMAP_GETPORT 3 static struct rpc_procinfo pmap_procedures[]; -static struct rpc_clnt * pmap_create(char *, struct sockaddr_in *, int); +static struct rpc_clnt * pmap_create(char *, struct sockaddr_in *, int, int); static void pmap_getport_done(struct rpc_task *); static struct rpc_program pmap_program; static DEFINE_SPINLOCK(pmap_lock); @@ -65,7 +65,7 @@ rpc_getport(struct rpc_task *task, struc map->pm_binding = 1; spin_unlock(&pmap_lock); - pmap_clnt = pmap_create(clnt->cl_server, sap, map->pm_prot); + pmap_clnt = pmap_create(clnt->cl_server, sap, map->pm_prot, 0); if (IS_ERR(pmap_clnt)) { task->tk_status = PTR_ERR(pmap_clnt); goto bailout; @@ -112,7 +112,7 @@ rpc_getport_external(struct sockaddr_in NIPQUAD(sin->sin_addr.s_addr), prog, vers, prot); sprintf(hostname, "%u.%u.%u.%u", NIPQUAD(sin->sin_addr.s_addr)); - pmap_clnt = pmap_create(hostname, sin, prot); + pmap_clnt = pmap_create(hostname, sin, prot, 0); if (IS_ERR(pmap_clnt)) return PTR_ERR(pmap_clnt); @@ -171,7 +171,7 @@ rpc_register(u32 prog, u32 vers, int pro sin.sin_family = AF_INET; sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); - pmap_clnt = pmap_create("localhost", &sin, IPPROTO_UDP); + pmap_clnt = pmap_create("localhost", &sin, IPPROTO_UDP, 1); if (IS_ERR(pmap_clnt)) { error = PTR_ERR(pmap_clnt); dprintk("RPC: couldn't create pmap client. Error = %d\n", error); @@ -198,7 +198,7 @@ rpc_register(u32 prog, u32 vers, int pro } static struct rpc_clnt * -pmap_create(char *hostname, struct sockaddr_in *srvaddr, int proto) +pmap_create(char *hostname, struct sockaddr_in *srvaddr, int proto, int privileged) { struct rpc_xprt *xprt; struct rpc_clnt *clnt; @@ -208,6 +208,8 @@ pmap_create(char *hostname, struct socka if (IS_ERR(xprt)) return (struct rpc_clnt *)xprt; xprt->addr.sin_port = htons(RPC_PMAP_PORT); + if (!privileged) + xprt->resvport = 0; /* printk("pmap: create clnt\n"); */ clnt = rpc_new_client(xprt, hostname, Index: linux-2.6/net/sunrpc/socklib.c =================================================================== --- /dev/null +++ linux-2.6/net/sunrpc/socklib.c @@ -0,0 +1,175 @@ +/* + * linux/net/sunrpc/socklib.c + * + * Common socket helper routines for RPC client and server + * + * Copyright (C) 1995, 1996 Olaf Kirch + */ + +#include +#include +#include +#include + + +/** + * skb_read_bits - copy some data bits from skb to internal buffer + * @desc: sk_buff copy helper + * @to: copy destination + * @len: number of bytes to copy + * + * Possibly called several times to iterate over an sk_buff and copy + * data out of it. + */ +static size_t skb_read_bits(skb_reader_t *desc, void *to, size_t len) +{ + if (len > desc->count) + len = desc->count; + if (skb_copy_bits(desc->skb, desc->offset, to, len)) + return 0; + desc->count -= len; + desc->offset += len; + return len; +} + +/** + * skb_read_and_csum_bits - copy and checksum from skb to buffer + * @desc: sk_buff copy helper + * @to: copy destination + * @len: number of bytes to copy + * + * Same as skb_read_bits, but calculate a checksum at the same time. + */ +static size_t skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len) +{ + unsigned int csum2, pos; + + if (len > desc->count) + len = desc->count; + pos = desc->offset; + csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0); + desc->csum = csum_block_add(desc->csum, csum2, pos); + desc->count -= len; + desc->offset += len; + return len; +} + +/** + * xdr_partial_copy_from_skb - copy data out of an skb + * @xdr: target XDR buffer + * @base: starting offset + * @desc: sk_buff copy helper + * @copy_actor: virtual method for copying data + * + */ +ssize_t xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base, skb_reader_t *desc, skb_read_actor_t copy_actor) +{ + struct page **ppage = xdr->pages; + unsigned int len, pglen = xdr->page_len; + ssize_t copied = 0; + int ret; + + len = xdr->head[0].iov_len; + if (base < len) { + len -= base; + ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len); + copied += ret; + if (ret != len || !desc->count) + goto out; + base = 0; + } else + base -= len; + + if (unlikely(pglen == 0)) + goto copy_tail; + if (unlikely(base >= pglen)) { + base -= pglen; + goto copy_tail; + } + if (base || xdr->page_base) { + pglen -= base; + base += xdr->page_base; + ppage += base >> PAGE_CACHE_SHIFT; + base &= ~PAGE_CACHE_MASK; + } + do { + char *kaddr; + + /* ACL likes to be lazy in allocating pages - ACLs + * are small by default but can get huge. */ + if (unlikely(*ppage == NULL)) { + *ppage = alloc_page(GFP_ATOMIC); + if (unlikely(*ppage == NULL)) { + if (copied == 0) + copied = -ENOMEM; + goto out; + } + } + + len = PAGE_CACHE_SIZE; + kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA); + if (base) { + len -= base; + if (pglen < len) + len = pglen; + ret = copy_actor(desc, kaddr + base, len); + base = 0; + } else { + if (pglen < len) + len = pglen; + ret = copy_actor(desc, kaddr, len); + } + flush_dcache_page(*ppage); + kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA); + copied += ret; + if (ret != len || !desc->count) + goto out; + ppage++; + } while ((pglen -= len) != 0); +copy_tail: + len = xdr->tail[0].iov_len; + if (base < len) + copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base); +out: + return copied; +} + +/** + * csum_partial_copy_to_xdr - checksum and copy data + * @xdr: target XDR buffer + * @skb: source skb + * + * We have set things up such that we perform the checksum of the UDP + * packet in parallel with the copies into the RPC client iovec. -DaveM + */ +int csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) +{ + skb_reader_t desc; + + desc.skb = skb; + desc.offset = sizeof(struct udphdr); + desc.count = skb->len - desc.offset; + + if (skb->ip_summed == CHECKSUM_UNNECESSARY) + goto no_checksum; + + desc.csum = csum_partial(skb->data, desc.offset, skb->csum); + if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0) + return -1; + if (desc.offset != skb->len) { + unsigned int csum2; + csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0); + desc.csum = csum_block_add(desc.csum, csum2, desc.offset); + } + if (desc.count) + return -1; + if ((unsigned short)csum_fold(desc.csum)) + return -1; + return 0; +no_checksum: + if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0) + return -1; + if (desc.count) + return -1; + return 0; +} Index: linux-2.6/net/sunrpc/sunrpc_syms.c =================================================================== --- linux-2.6.orig/net/sunrpc/sunrpc_syms.c +++ linux-2.6/net/sunrpc/sunrpc_syms.c @@ -10,7 +10,6 @@ #include #include -#include #include #include #include Index: linux-2.6/net/sunrpc/svcsock.c =================================================================== --- linux-2.6.orig/net/sunrpc/svcsock.c +++ linux-2.6/net/sunrpc/svcsock.c @@ -548,9 +548,6 @@ svc_write_space(struct sock *sk) /* * Receive a datagram from a UDP socket. */ -extern int -csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb); - static int svc_udp_recvfrom(struct svc_rqst *rqstp) { Index: linux-2.6/net/sunrpc/sysctl.c =================================================================== --- linux-2.6.orig/net/sunrpc/sysctl.c +++ linux-2.6/net/sunrpc/sysctl.c @@ -119,8 +119,18 @@ done: return 0; } +unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; +unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; +unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; +EXPORT_SYMBOL(xprt_min_resvport); +unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; +EXPORT_SYMBOL(xprt_max_resvport); + + static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; +static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; +static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; static ctl_table debug_table[] = { { @@ -177,6 +187,28 @@ static ctl_table debug_table[] = { .extra1 = &min_slot_table_size, .extra2 = &max_slot_table_size }, + { + .ctl_name = CTL_MIN_RESVPORT, + .procname = "min_resvport", + .data = &xprt_min_resvport, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .strategy = &sysctl_intvec, + .extra1 = &xprt_min_resvport_limit, + .extra2 = &xprt_max_resvport_limit + }, + { + .ctl_name = CTL_MAX_RESVPORT, + .procname = "max_resvport", + .data = &xprt_max_resvport, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .strategy = &sysctl_intvec, + .extra1 = &xprt_min_resvport_limit, + .extra2 = &xprt_max_resvport_limit + }, { .ctl_name = 0 } }; Index: linux-2.6/net/sunrpc/xdr.c =================================================================== --- linux-2.6.orig/net/sunrpc/xdr.c +++ linux-2.6/net/sunrpc/xdr.c @@ -6,15 +6,12 @@ * Copyright (C) 1995, 1996 Olaf Kirch */ +#include #include -#include #include #include #include #include -#include -#include -#include #include #include @@ -176,178 +173,6 @@ xdr_inline_pages(struct xdr_buf *xdr, un xdr->buflen += len; } -ssize_t -xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base, - skb_reader_t *desc, - skb_read_actor_t copy_actor) -{ - struct page **ppage = xdr->pages; - unsigned int len, pglen = xdr->page_len; - ssize_t copied = 0; - int ret; - - len = xdr->head[0].iov_len; - if (base < len) { - len -= base; - ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len); - copied += ret; - if (ret != len || !desc->count) - goto out; - base = 0; - } else - base -= len; - - if (pglen == 0) - goto copy_tail; - if (base >= pglen) { - base -= pglen; - goto copy_tail; - } - if (base || xdr->page_base) { - pglen -= base; - base += xdr->page_base; - ppage += base >> PAGE_CACHE_SHIFT; - base &= ~PAGE_CACHE_MASK; - } - do { - char *kaddr; - - /* ACL likes to be lazy in allocating pages - ACLs - * are small by default but can get huge. */ - if (unlikely(*ppage == NULL)) { - *ppage = alloc_page(GFP_ATOMIC); - if (unlikely(*ppage == NULL)) { - if (copied == 0) - copied = -ENOMEM; - goto out; - } - } - - len = PAGE_CACHE_SIZE; - kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA); - if (base) { - len -= base; - if (pglen < len) - len = pglen; - ret = copy_actor(desc, kaddr + base, len); - base = 0; - } else { - if (pglen < len) - len = pglen; - ret = copy_actor(desc, kaddr, len); - } - flush_dcache_page(*ppage); - kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA); - copied += ret; - if (ret != len || !desc->count) - goto out; - ppage++; - } while ((pglen -= len) != 0); -copy_tail: - len = xdr->tail[0].iov_len; - if (base < len) - copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base); -out: - return copied; -} - - -int -xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, - struct xdr_buf *xdr, unsigned int base, int msgflags) -{ - struct page **ppage = xdr->pages; - unsigned int len, pglen = xdr->page_len; - int err, ret = 0; - ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int); - - len = xdr->head[0].iov_len; - if (base < len || (addr != NULL && base == 0)) { - struct kvec iov = { - .iov_base = xdr->head[0].iov_base + base, - .iov_len = len - base, - }; - struct msghdr msg = { - .msg_name = addr, - .msg_namelen = addrlen, - .msg_flags = msgflags, - }; - if (xdr->len > len) - msg.msg_flags |= MSG_MORE; - - if (iov.iov_len != 0) - err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); - else - err = kernel_sendmsg(sock, &msg, NULL, 0, 0); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - if (err != iov.iov_len) - goto out; - base = 0; - } else - base -= len; - - if (pglen == 0) - goto copy_tail; - if (base >= pglen) { - base -= pglen; - goto copy_tail; - } - if (base || xdr->page_base) { - pglen -= base; - base += xdr->page_base; - ppage += base >> PAGE_CACHE_SHIFT; - base &= ~PAGE_CACHE_MASK; - } - - sendpage = sock->ops->sendpage ? : sock_no_sendpage; - do { - int flags = msgflags; - - len = PAGE_CACHE_SIZE; - if (base) - len -= base; - if (pglen < len) - len = pglen; - - if (pglen != len || xdr->tail[0].iov_len != 0) - flags |= MSG_MORE; - - /* Hmm... We might be dealing with highmem pages */ - if (PageHighMem(*ppage)) - sendpage = sock_no_sendpage; - err = sendpage(sock, *ppage, base, len, flags); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - if (err != len) - goto out; - base = 0; - ppage++; - } while ((pglen -= len) != 0); -copy_tail: - len = xdr->tail[0].iov_len; - if (base < len) { - struct kvec iov = { - .iov_base = xdr->tail[0].iov_base + base, - .iov_len = len - base, - }; - struct msghdr msg = { - .msg_flags = msgflags, - }; - err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - } -out: - return ret; -} - /* * Helper routines for doing 'memmove' like operations on a struct xdr_buf Index: linux-2.6/net/sunrpc/xprt.c =================================================================== --- linux-2.6.orig/net/sunrpc/xprt.c +++ linux-2.6/net/sunrpc/xprt.c @@ -10,12 +10,12 @@ * one is available. Otherwise, it sleeps on the backlog queue * (xprt_reserve). * - Next, the caller puts together the RPC message, stuffs it into - * the request struct, and calls xprt_call(). - * - xprt_call transmits the message and installs the caller on the - * socket's wait list. At the same time, it installs a timer that + * the request struct, and calls xprt_transmit(). + * - xprt_transmit sends the message and installs the caller on the + * transport's wait list. At the same time, it installs a timer that * is run after the packet's timeout has expired. * - When a packet arrives, the data_ready handler walks the list of - * pending requests for that socket. If a matching XID is found, the + * pending requests for that transport. If a matching XID is found, the * caller is woken up, and the timer removed. * - When no reply arrives within the timeout interval, the timer is * fired by the kernel and runs xprt_timer(). It either adjusts the @@ -33,36 +33,17 @@ * * Copyright (C) 1995-1997, Olaf Kirch * - * TCP callback races fixes (C) 1998 Red Hat Software - * TCP send fixes (C) 1998 Red Hat Software - * TCP NFS related read + write fixes - * (C) 1999 Dave Airlie, University of Limerick, Ireland - * - * Rewrite of larges part of the code in order to stabilize TCP stuff. - * Fix behaviour when socket buffer is full. - * (C) 1999 Trond Myklebust + * Transport switch API copyright (C) 2005, Chuck Lever */ +#include + #include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include +#include #include #include -#include -#include -#include -#include +#include /* * Local variables @@ -73,81 +54,90 @@ # define RPCDBG_FACILITY RPCDBG_XPRT #endif -#define XPRT_MAX_BACKOFF (8) -#define XPRT_IDLE_TIMEOUT (5*60*HZ) -#define XPRT_MAX_RESVPORT (800) - /* * Local functions */ static void xprt_request_init(struct rpc_task *, struct rpc_xprt *); static inline void do_xprt_reserve(struct rpc_task *); -static void xprt_disconnect(struct rpc_xprt *); static void xprt_connect_status(struct rpc_task *task); -static struct rpc_xprt * xprt_setup(int proto, struct sockaddr_in *ap, - struct rpc_timeout *to); -static struct socket *xprt_create_socket(struct rpc_xprt *, int, int); -static void xprt_bind_socket(struct rpc_xprt *, struct socket *); static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *); -static int xprt_clear_backlog(struct rpc_xprt *xprt); - -#ifdef RPC_DEBUG_DATA /* - * Print the buffer contents (first 128 bytes only--just enough for - * diropres return). + * The transport code maintains an estimate on the maximum number of out- + * standing RPC requests, using a smoothed version of the congestion + * avoidance implemented in 44BSD. This is basically the Van Jacobson + * congestion algorithm: If a retransmit occurs, the congestion window is + * halved; otherwise, it is incremented by 1/cwnd when + * + * - a reply is received and + * - a full number of requests are outstanding and + * - the congestion window hasn't been updated recently. + */ +#define RPC_CWNDSHIFT (8U) +#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT) +#define RPC_INITCWND RPC_CWNDSCALE +#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT) + +#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd) + +/** + * xprt_reserve_xprt - serialize write access to transports + * @task: task that is requesting access to the transport + * + * This prevents mixing the payload of separate requests, and prevents + * transport connects from colliding with writes. No congestion control + * is provided. */ -static void -xprt_pktdump(char *msg, u32 *packet, unsigned int count) +int xprt_reserve_xprt(struct rpc_task *task) { - u8 *buf = (u8 *) packet; - int j; + struct rpc_xprt *xprt = task->tk_xprt; + struct rpc_rqst *req = task->tk_rqstp; - dprintk("RPC: %s\n", msg); - for (j = 0; j < count && j < 128; j += 4) { - if (!(j & 31)) { - if (j) - dprintk("\n"); - dprintk("0x%04x ", j); - } - dprintk("%02x%02x%02x%02x ", - buf[j], buf[j+1], buf[j+2], buf[j+3]); + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) { + if (task == xprt->snd_task) + return 1; + if (task == NULL) + return 0; + goto out_sleep; } - dprintk("\n"); -} -#else -static inline void -xprt_pktdump(char *msg, u32 *packet, unsigned int count) -{ - /* NOP */ -} -#endif + xprt->snd_task = task; + if (req) { + req->rq_bytes_sent = 0; + req->rq_ntrans++; + } + return 1; -/* - * Look up RPC transport given an INET socket - */ -static inline struct rpc_xprt * -xprt_from_sock(struct sock *sk) -{ - return (struct rpc_xprt *) sk->sk_user_data; +out_sleep: + dprintk("RPC: %4d failed to lock transport %p\n", + task->tk_pid, xprt); + task->tk_timeout = 0; + task->tk_status = -EAGAIN; + if (req && req->rq_ntrans) + rpc_sleep_on(&xprt->resend, task, NULL, NULL); + else + rpc_sleep_on(&xprt->sending, task, NULL, NULL); + return 0; } /* - * Serialize write access to sockets, in order to prevent different - * requests from interfering with each other. - * Also prevents TCP socket connects from colliding with writes. + * xprt_reserve_xprt_cong - serialize write access to transports + * @task: task that is requesting access to the transport + * + * Same as xprt_reserve_xprt, but Van Jacobson congestion control is + * integrated into the decision of whether a request is allowed to be + * woken up and given access to the transport. */ -static int -__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) +int xprt_reserve_xprt_cong(struct rpc_task *task) { + struct rpc_xprt *xprt = task->tk_xprt; struct rpc_rqst *req = task->tk_rqstp; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) { + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) { if (task == xprt->snd_task) return 1; goto out_sleep; } - if (xprt->nocong || __xprt_get_cong(xprt, task)) { + if (__xprt_get_cong(xprt, task)) { xprt->snd_task = task; if (req) { req->rq_bytes_sent = 0; @@ -156,10 +146,10 @@ __xprt_lock_write(struct rpc_xprt *xprt, return 1; } smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); out_sleep: - dprintk("RPC: %4d failed to lock socket %p\n", task->tk_pid, xprt); + dprintk("RPC: %4d failed to lock transport %p\n", task->tk_pid, xprt); task->tk_timeout = 0; task->tk_status = -EAGAIN; if (req && req->rq_ntrans) @@ -169,26 +159,52 @@ out_sleep: return 0; } -static inline int -xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) +static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) { int retval; - spin_lock_bh(&xprt->sock_lock); - retval = __xprt_lock_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); + retval = xprt->ops->reserve_xprt(task); + spin_unlock_bh(&xprt->transport_lock); return retval; } +static void __xprt_lock_write_next(struct rpc_xprt *xprt) +{ + struct rpc_task *task; + struct rpc_rqst *req; -static void -__xprt_lock_write_next(struct rpc_xprt *xprt) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) + return; + + task = rpc_wake_up_next(&xprt->resend); + if (!task) { + task = rpc_wake_up_next(&xprt->sending); + if (!task) + goto out_unlock; + } + + req = task->tk_rqstp; + xprt->snd_task = task; + if (req) { + req->rq_bytes_sent = 0; + req->rq_ntrans++; + } + return; + +out_unlock: + smp_mb__before_clear_bit(); + clear_bit(XPRT_LOCKED, &xprt->state); + smp_mb__after_clear_bit(); +} + +static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt) { struct rpc_task *task; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) return; - if (!xprt->nocong && RPCXPRT_CONGESTED(xprt)) + if (RPCXPRT_CONGESTED(xprt)) goto out_unlock; task = rpc_wake_up_next(&xprt->resend); if (!task) { @@ -196,7 +212,7 @@ __xprt_lock_write_next(struct rpc_xprt * if (!task) goto out_unlock; } - if (xprt->nocong || __xprt_get_cong(xprt, task)) { + if (__xprt_get_cong(xprt, task)) { struct rpc_rqst *req = task->tk_rqstp; xprt->snd_task = task; if (req) { @@ -207,87 +223,52 @@ __xprt_lock_write_next(struct rpc_xprt * } out_unlock: smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); } -/* - * Releases the socket for use by other requests. +/** + * xprt_release_xprt - allow other requests to use a transport + * @xprt: transport with other tasks potentially waiting + * @task: task that is releasing access to the transport + * + * Note that "task" can be NULL. No congestion control is provided. */ -static void -__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) +void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) { if (xprt->snd_task == task) { xprt->snd_task = NULL; smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); __xprt_lock_write_next(xprt); } } -static inline void -xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) -{ - spin_lock_bh(&xprt->sock_lock); - __xprt_release_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); -} - -/* - * Write data to socket. +/** + * xprt_release_xprt_cong - allow other requests to use a transport + * @xprt: transport with other tasks potentially waiting + * @task: task that is releasing access to the transport + * + * Note that "task" can be NULL. Another task is awoken to use the + * transport if the transport's congestion window allows it. */ -static inline int -xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req) +void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task) { - struct socket *sock = xprt->sock; - struct xdr_buf *xdr = &req->rq_snd_buf; - struct sockaddr *addr = NULL; - int addrlen = 0; - unsigned int skip; - int result; - - if (!sock) - return -ENOTCONN; - - xprt_pktdump("packet data:", - req->rq_svec->iov_base, - req->rq_svec->iov_len); - - /* For UDP, we need to provide an address */ - if (!xprt->stream) { - addr = (struct sockaddr *) &xprt->addr; - addrlen = sizeof(xprt->addr); + if (xprt->snd_task == task) { + xprt->snd_task = NULL; + smp_mb__before_clear_bit(); + clear_bit(XPRT_LOCKED, &xprt->state); + smp_mb__after_clear_bit(); + __xprt_lock_write_next_cong(xprt); } - /* Dont repeat bytes */ - skip = req->rq_bytes_sent; - - clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); - result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT); - - dprintk("RPC: xprt_sendmsg(%d) = %d\n", xdr->len - skip, result); - - if (result >= 0) - return result; +} - switch (result) { - case -ECONNREFUSED: - /* When the server has died, an ICMP port unreachable message - * prompts ECONNREFUSED. - */ - case -EAGAIN: - break; - case -ECONNRESET: - case -ENOTCONN: - case -EPIPE: - /* connection broken */ - if (xprt->stream) - result = -ENOTCONN; - break; - default: - printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result); - } - return result; +static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) +{ + spin_lock_bh(&xprt->transport_lock); + xprt->ops->release_xprt(xprt, task); + spin_unlock_bh(&xprt->transport_lock); } /* @@ -321,26 +302,40 @@ __xprt_put_cong(struct rpc_xprt *xprt, s return; req->rq_cong = 0; xprt->cong -= RPC_CWNDSCALE; - __xprt_lock_write_next(xprt); + __xprt_lock_write_next_cong(xprt); } -/* - * Adjust RPC congestion window +/** + * xprt_release_rqst_cong - housekeeping when request is complete + * @task: RPC request that recently completed + * + * Useful for transports that require congestion control. + */ +void xprt_release_rqst_cong(struct rpc_task *task) +{ + __xprt_put_cong(task->tk_xprt, task->tk_rqstp); +} + +/** + * xprt_adjust_cwnd - adjust transport congestion window + * @task: recently completed RPC request used to adjust window + * @result: result code of completed RPC request + * * We use a time-smoothed congestion estimator to avoid heavy oscillation. */ -static void -xprt_adjust_cwnd(struct rpc_xprt *xprt, int result) +void xprt_adjust_cwnd(struct rpc_task *task, int result) { - unsigned long cwnd; + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = task->tk_xprt; + unsigned long cwnd = xprt->cwnd; - cwnd = xprt->cwnd; if (result >= 0 && cwnd <= xprt->cong) { /* The (cwnd >> 1) term makes sure * the result gets rounded properly. */ cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd; if (cwnd > RPC_MAXCWND(xprt)) cwnd = RPC_MAXCWND(xprt); - __xprt_lock_write_next(xprt); + __xprt_lock_write_next_cong(xprt); } else if (result == -ETIMEDOUT) { cwnd >>= 1; if (cwnd < RPC_CWNDSCALE) @@ -349,11 +344,89 @@ xprt_adjust_cwnd(struct rpc_xprt *xprt, dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n", xprt->cong, xprt->cwnd, cwnd); xprt->cwnd = cwnd; + __xprt_put_cong(xprt, req); +} + +/** + * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue + * @xprt: transport with waiting tasks + * @status: result code to plant in each task before waking it + * + */ +void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status) +{ + if (status < 0) + rpc_wake_up_status(&xprt->pending, status); + else + rpc_wake_up(&xprt->pending); +} + +/** + * xprt_wait_for_buffer_space - wait for transport output buffer to clear + * @task: task to be put to sleep + * + */ +void xprt_wait_for_buffer_space(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + + task->tk_timeout = req->rq_timeout; + rpc_sleep_on(&xprt->pending, task, NULL, NULL); +} + +/** + * xprt_write_space - wake the task waiting for transport output buffer space + * @xprt: transport with waiting tasks + * + * Can be called in a soft IRQ context, so xprt_write_space never sleeps. + */ +void xprt_write_space(struct rpc_xprt *xprt) +{ + if (unlikely(xprt->shutdown)) + return; + + spin_lock_bh(&xprt->transport_lock); + if (xprt->snd_task) { + dprintk("RPC: write space: waking waiting task on xprt %p\n", + xprt); + rpc_wake_up_task(xprt->snd_task); + } + spin_unlock_bh(&xprt->transport_lock); +} + +/** + * xprt_set_retrans_timeout_def - set a request's retransmit timeout + * @task: task whose timeout is to be set + * + * Set a request's retransmit timeout based on the transport's + * default timeout parameters. Used by transports that don't adjust + * the retransmit timeout based on round-trip time estimation. + */ +void xprt_set_retrans_timeout_def(struct rpc_task *task) +{ + task->tk_timeout = task->tk_rqstp->rq_timeout; } /* - * Reset the major timeout value + * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout + * @task: task whose timeout is to be set + * + * Set a request's retransmit timeout using the RTT estimator. */ +void xprt_set_retrans_timeout_rtt(struct rpc_task *task) +{ + int timer = task->tk_msg.rpc_proc->p_timer; + struct rpc_rtt *rtt = task->tk_client->cl_rtt; + struct rpc_rqst *req = task->tk_rqstp; + unsigned long max_timeout = req->rq_xprt->timeout.to_maxval; + + task->tk_timeout = rpc_calc_rto(rtt, timer); + task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries; + if (task->tk_timeout > max_timeout || task->tk_timeout == 0) + task->tk_timeout = max_timeout; +} + static void xprt_reset_majortimeo(struct rpc_rqst *req) { struct rpc_timeout *to = &req->rq_xprt->timeout; @@ -368,8 +441,10 @@ static void xprt_reset_majortimeo(struct req->rq_majortimeo += jiffies; } -/* - * Adjust timeout values etc for next retransmit +/** + * xprt_adjust_timeout - adjust timeout values for next retransmit + * @req: RPC request containing parameters to use for the adjustment + * */ int xprt_adjust_timeout(struct rpc_rqst *req) { @@ -391,9 +466,9 @@ int xprt_adjust_timeout(struct rpc_rqst req->rq_retries = 0; xprt_reset_majortimeo(req); /* Reset the RTT counters == "slow start" */ - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval); - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); pprintk("RPC: %lu timeout\n", jiffies); status = -ETIMEDOUT; } @@ -405,133 +480,52 @@ int xprt_adjust_timeout(struct rpc_rqst return status; } -/* - * Close down a transport socket - */ -static void -xprt_close(struct rpc_xprt *xprt) -{ - struct socket *sock = xprt->sock; - struct sock *sk = xprt->inet; - - if (!sk) - return; - - write_lock_bh(&sk->sk_callback_lock); - xprt->inet = NULL; - xprt->sock = NULL; - - sk->sk_user_data = NULL; - sk->sk_data_ready = xprt->old_data_ready; - sk->sk_state_change = xprt->old_state_change; - sk->sk_write_space = xprt->old_write_space; - write_unlock_bh(&sk->sk_callback_lock); - - sk->sk_no_check = 0; - - sock_release(sock); -} - -static void -xprt_socket_autoclose(void *args) +static void xprt_autoclose(void *args) { struct rpc_xprt *xprt = (struct rpc_xprt *)args; xprt_disconnect(xprt); - xprt_close(xprt); + xprt->ops->close(xprt); xprt_release_write(xprt, NULL); } -/* - * Mark a transport as disconnected +/** + * xprt_disconnect - mark a transport as disconnected + * @xprt: transport to flag for disconnect + * */ -static void -xprt_disconnect(struct rpc_xprt *xprt) +void xprt_disconnect(struct rpc_xprt *xprt) { dprintk("RPC: disconnected transport %p\n", xprt); - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); xprt_clear_connected(xprt); - rpc_wake_up_status(&xprt->pending, -ENOTCONN); - spin_unlock_bh(&xprt->sock_lock); + xprt_wake_pending_tasks(xprt, -ENOTCONN); + spin_unlock_bh(&xprt->transport_lock); } -/* - * Used to allow disconnection when we've been idle - */ static void xprt_init_autodisconnect(unsigned long data) { struct rpc_xprt *xprt = (struct rpc_xprt *)data; - spin_lock(&xprt->sock_lock); + spin_lock(&xprt->transport_lock); if (!list_empty(&xprt->recv) || xprt->shutdown) goto out_abort; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) goto out_abort; - spin_unlock(&xprt->sock_lock); - /* Let keventd close the socket */ - if (test_bit(XPRT_CONNECTING, &xprt->sockstate) != 0) + spin_unlock(&xprt->transport_lock); + if (xprt_connecting(xprt)) xprt_release_write(xprt, NULL); else schedule_work(&xprt->task_cleanup); return; out_abort: - spin_unlock(&xprt->sock_lock); + spin_unlock(&xprt->transport_lock); } -static void xprt_socket_connect(void *args) -{ - struct rpc_xprt *xprt = (struct rpc_xprt *)args; - struct socket *sock = xprt->sock; - int status = -EIO; - - if (xprt->shutdown || xprt->addr.sin_port == 0) - goto out; - - /* - * Start by resetting any existing state - */ - xprt_close(xprt); - sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport); - if (sock == NULL) { - /* couldn't create socket or bind to reserved port; - * this is likely a permanent error, so cause an abort */ - goto out; - } - xprt_bind_socket(xprt, sock); - xprt_sock_setbufsize(xprt); - - status = 0; - if (!xprt->stream) - goto out; - - /* - * Tell the socket layer to start connecting... - */ - status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr, - sizeof(xprt->addr), O_NONBLOCK); - dprintk("RPC: %p connect status %d connected %d sock state %d\n", - xprt, -status, xprt_connected(xprt), sock->sk->sk_state); - if (status < 0) { - switch (status) { - case -EINPROGRESS: - case -EALREADY: - goto out_clear; - } - } -out: - if (status < 0) - rpc_wake_up_status(&xprt->pending, status); - else - rpc_wake_up(&xprt->pending); -out_clear: - smp_mb__before_clear_bit(); - clear_bit(XPRT_CONNECTING, &xprt->sockstate); - smp_mb__after_clear_bit(); -} - -/* - * Attempt to connect a TCP socket. +/** + * xprt_connect - schedule a transport connect operation + * @task: RPC task that is requesting the connect * */ void xprt_connect(struct rpc_task *task) @@ -552,37 +546,19 @@ void xprt_connect(struct rpc_task *task) if (!xprt_lock_write(xprt, task)) return; if (xprt_connected(xprt)) - goto out_write; - - if (task->tk_rqstp) - task->tk_rqstp->rq_bytes_sent = 0; - - task->tk_timeout = RPC_CONNECT_TIMEOUT; - rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL); - if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) { - /* Note: if we are here due to a dropped connection - * we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ - * seconds - */ - if (xprt->sock != NULL) - schedule_delayed_work(&xprt->sock_connect, - RPC_REESTABLISH_TIMEOUT); - else { - schedule_work(&xprt->sock_connect); - if (!RPC_IS_ASYNC(task)) - flush_scheduled_work(); - } + xprt_release_write(xprt, task); + else { + if (task->tk_rqstp) + task->tk_rqstp->rq_bytes_sent = 0; + + task->tk_timeout = xprt->connect_timeout; + rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL); + xprt->ops->connect(task); } return; - out_write: - xprt_release_write(xprt, task); } -/* - * We arrive here when awoken from waiting on connection establishment. - */ -static void -xprt_connect_status(struct rpc_task *task) +static void xprt_connect_status(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; @@ -592,31 +568,42 @@ xprt_connect_status(struct rpc_task *tas return; } - /* if soft mounted, just cause this RPC to fail */ - if (RPC_IS_SOFT(task)) - task->tk_status = -EIO; - switch (task->tk_status) { case -ECONNREFUSED: case -ECONNRESET: + dprintk("RPC: %4d xprt_connect_status: server %s refused connection\n", + task->tk_pid, task->tk_client->cl_server); + break; case -ENOTCONN: - return; + dprintk("RPC: %4d xprt_connect_status: connection broken\n", + task->tk_pid); + break; case -ETIMEDOUT: - dprintk("RPC: %4d xprt_connect_status: timed out\n", + dprintk("RPC: %4d xprt_connect_status: connect attempt timed out\n", task->tk_pid); break; default: - printk(KERN_ERR "RPC: error %d connecting to server %s\n", - -task->tk_status, task->tk_client->cl_server); + dprintk("RPC: %4d xprt_connect_status: error %d connecting to server %s\n", + task->tk_pid, -task->tk_status, task->tk_client->cl_server); + xprt_release_write(xprt, task); + task->tk_status = -EIO; + return; + } + + /* if soft mounted, just cause this RPC to fail */ + if (RPC_IS_SOFT(task)) { + xprt_release_write(xprt, task); + task->tk_status = -EIO; } - xprt_release_write(xprt, task); } -/* - * Look up the RPC request corresponding to a reply, and then lock it. +/** + * xprt_lookup_rqst - find an RPC request corresponding to an XID + * @xprt: transport on which the original request was transmitted + * @xid: RPC XID of incoming reply + * */ -static inline struct rpc_rqst * -xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid) +struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid) { struct list_head *pos; struct rpc_rqst *req = NULL; @@ -631,556 +618,68 @@ xprt_lookup_rqst(struct rpc_xprt *xprt, return req; } -/* - * Complete reply received. - * The TCP code relies on us to remove the request from xprt->pending. - */ -static void -xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied) -{ - struct rpc_task *task = req->rq_task; - struct rpc_clnt *clnt = task->tk_client; - - /* Adjust congestion window */ - if (!xprt->nocong) { - unsigned timer = task->tk_msg.rpc_proc->p_timer; - xprt_adjust_cwnd(xprt, copied); - __xprt_put_cong(xprt, req); - if (timer) { - if (req->rq_ntrans == 1) - rpc_update_rtt(clnt->cl_rtt, timer, - (long)jiffies - req->rq_xtime); - rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1); - } - } - -#ifdef RPC_PROFILE - /* Profile only reads for now */ - if (copied > 1024) { - static unsigned long nextstat; - static unsigned long pkt_rtt, pkt_len, pkt_cnt; - - pkt_cnt++; - pkt_len += req->rq_slen + copied; - pkt_rtt += jiffies - req->rq_xtime; - if (time_before(nextstat, jiffies)) { - printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd); - printk("RPC: %ld %ld %ld %ld stat\n", - jiffies, pkt_cnt, pkt_len, pkt_rtt); - pkt_rtt = pkt_len = pkt_cnt = 0; - nextstat = jiffies + 5 * HZ; - } - } -#endif - - dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied); - list_del_init(&req->rq_list); - req->rq_received = req->rq_private_buf.len = copied; - - /* ... and wake up the process. */ - rpc_wake_up_task(task); - return; -} - -static size_t -skb_read_bits(skb_reader_t *desc, void *to, size_t len) -{ - if (len > desc->count) - len = desc->count; - if (skb_copy_bits(desc->skb, desc->offset, to, len)) - return 0; - desc->count -= len; - desc->offset += len; - return len; -} - -static size_t -skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len) -{ - unsigned int csum2, pos; - - if (len > desc->count) - len = desc->count; - pos = desc->offset; - csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0); - desc->csum = csum_block_add(desc->csum, csum2, pos); - desc->count -= len; - desc->offset += len; - return len; -} - -/* - * We have set things up such that we perform the checksum of the UDP - * packet in parallel with the copies into the RPC client iovec. -DaveM - */ -int -csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) -{ - skb_reader_t desc; - - desc.skb = skb; - desc.offset = sizeof(struct udphdr); - desc.count = skb->len - desc.offset; - - if (skb->ip_summed == CHECKSUM_UNNECESSARY) - goto no_checksum; - - desc.csum = csum_partial(skb->data, desc.offset, skb->csum); - if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0) - return -1; - if (desc.offset != skb->len) { - unsigned int csum2; - csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0); - desc.csum = csum_block_add(desc.csum, csum2, desc.offset); - } - if (desc.count) - return -1; - if ((unsigned short)csum_fold(desc.csum)) - return -1; - return 0; -no_checksum: - if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0) - return -1; - if (desc.count) - return -1; - return 0; -} - -/* - * Input handler for RPC replies. Called from a bottom half and hence - * atomic. - */ -static void -udp_data_ready(struct sock *sk, int len) -{ - struct rpc_task *task; - struct rpc_xprt *xprt; - struct rpc_rqst *rovr; - struct sk_buff *skb; - int err, repsize, copied; - u32 _xid, *xp; - - read_lock(&sk->sk_callback_lock); - dprintk("RPC: udp_data_ready...\n"); - if (!(xprt = xprt_from_sock(sk))) { - printk("RPC: udp_data_ready request not found!\n"); - goto out; - } - - dprintk("RPC: udp_data_ready client %p\n", xprt); - - if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) - goto out; - - if (xprt->shutdown) - goto dropit; - - repsize = skb->len - sizeof(struct udphdr); - if (repsize < 4) { - printk("RPC: impossible RPC reply size %d!\n", repsize); - goto dropit; - } - - /* Copy the XID from the skb... */ - xp = skb_header_pointer(skb, sizeof(struct udphdr), - sizeof(_xid), &_xid); - if (xp == NULL) - goto dropit; - - /* Look up and lock the request corresponding to the given XID */ - spin_lock(&xprt->sock_lock); - rovr = xprt_lookup_rqst(xprt, *xp); - if (!rovr) - goto out_unlock; - task = rovr->rq_task; - - dprintk("RPC: %4d received reply\n", task->tk_pid); - - if ((copied = rovr->rq_private_buf.buflen) > repsize) - copied = repsize; - - /* Suck it into the iovec, verify checksum if not done by hw. */ - if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) - goto out_unlock; - - /* Something worked... */ - dst_confirm(skb->dst); - - xprt_complete_rqst(xprt, rovr, copied); - - out_unlock: - spin_unlock(&xprt->sock_lock); - dropit: - skb_free_datagram(sk, skb); - out: - read_unlock(&sk->sk_callback_lock); -} - -/* - * Copy from an skb into memory and shrink the skb. - */ -static inline size_t -tcp_copy_data(skb_reader_t *desc, void *p, size_t len) -{ - if (len > desc->count) - len = desc->count; - if (skb_copy_bits(desc->skb, desc->offset, p, len)) { - dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n", - len, desc->count); - return 0; - } - desc->offset += len; - desc->count -= len; - dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n", - len, desc->count); - return len; -} - -/* - * TCP read fragment marker - */ -static inline void -tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len, used; - char *p; - - p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset; - len = sizeof(xprt->tcp_recm) - xprt->tcp_offset; - used = tcp_copy_data(desc, p, len); - xprt->tcp_offset += used; - if (used != len) - return; - xprt->tcp_reclen = ntohl(xprt->tcp_recm); - if (xprt->tcp_reclen & 0x80000000) - xprt->tcp_flags |= XPRT_LAST_FRAG; - else - xprt->tcp_flags &= ~XPRT_LAST_FRAG; - xprt->tcp_reclen &= 0x7fffffff; - xprt->tcp_flags &= ~XPRT_COPY_RECM; - xprt->tcp_offset = 0; - /* Sanity check of the record length */ - if (xprt->tcp_reclen < 4) { - printk(KERN_ERR "RPC: Invalid TCP record fragment length\n"); - xprt_disconnect(xprt); - } - dprintk("RPC: reading TCP record fragment of length %d\n", - xprt->tcp_reclen); -} - -static void -tcp_check_recm(struct rpc_xprt *xprt) -{ - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags); - if (xprt->tcp_offset == xprt->tcp_reclen) { - xprt->tcp_flags |= XPRT_COPY_RECM; - xprt->tcp_offset = 0; - if (xprt->tcp_flags & XPRT_LAST_FRAG) { - xprt->tcp_flags &= ~XPRT_COPY_DATA; - xprt->tcp_flags |= XPRT_COPY_XID; - xprt->tcp_copied = 0; - } - } -} - -/* - * TCP read xid - */ -static inline void -tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len, used; - char *p; - - len = sizeof(xprt->tcp_xid) - xprt->tcp_offset; - dprintk("RPC: reading XID (%Zu bytes)\n", len); - p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset; - used = tcp_copy_data(desc, p, len); - xprt->tcp_offset += used; - if (used != len) - return; - xprt->tcp_flags &= ~XPRT_COPY_XID; - xprt->tcp_flags |= XPRT_COPY_DATA; - xprt->tcp_copied = 4; - dprintk("RPC: reading reply for XID %08x\n", - ntohl(xprt->tcp_xid)); - tcp_check_recm(xprt); -} - -/* - * TCP read and complete request - */ -static inline void -tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - struct rpc_rqst *req; - struct xdr_buf *rcvbuf; - size_t len; - ssize_t r; - - /* Find and lock the request corresponding to this xid */ - spin_lock(&xprt->sock_lock); - req = xprt_lookup_rqst(xprt, xprt->tcp_xid); - if (!req) { - xprt->tcp_flags &= ~XPRT_COPY_DATA; - dprintk("RPC: XID %08x request not found!\n", - ntohl(xprt->tcp_xid)); - spin_unlock(&xprt->sock_lock); - return; - } - - rcvbuf = &req->rq_private_buf; - len = desc->count; - if (len > xprt->tcp_reclen - xprt->tcp_offset) { - skb_reader_t my_desc; - - len = xprt->tcp_reclen - xprt->tcp_offset; - memcpy(&my_desc, desc, sizeof(my_desc)); - my_desc.count = len; - r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, - &my_desc, tcp_copy_data); - desc->count -= r; - desc->offset += r; - } else - r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, - desc, tcp_copy_data); - - if (r > 0) { - xprt->tcp_copied += r; - xprt->tcp_offset += r; - } - if (r != len) { - /* Error when copying to the receive buffer, - * usually because we weren't able to allocate - * additional buffer pages. All we can do now - * is turn off XPRT_COPY_DATA, so the request - * will not receive any additional updates, - * and time out. - * Any remaining data from this record will - * be discarded. - */ - xprt->tcp_flags &= ~XPRT_COPY_DATA; - dprintk("RPC: XID %08x truncated request\n", - ntohl(xprt->tcp_xid)); - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); - goto out; - } - - dprintk("RPC: XID %08x read %Zd bytes\n", - ntohl(xprt->tcp_xid), r); - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); - - if (xprt->tcp_copied == req->rq_private_buf.buflen) - xprt->tcp_flags &= ~XPRT_COPY_DATA; - else if (xprt->tcp_offset == xprt->tcp_reclen) { - if (xprt->tcp_flags & XPRT_LAST_FRAG) - xprt->tcp_flags &= ~XPRT_COPY_DATA; - } - -out: - if (!(xprt->tcp_flags & XPRT_COPY_DATA)) { - dprintk("RPC: %4d received reply complete\n", - req->rq_task->tk_pid); - xprt_complete_rqst(xprt, req, xprt->tcp_copied); - } - spin_unlock(&xprt->sock_lock); - tcp_check_recm(xprt); -} - -/* - * TCP discard extra bytes from a short read - */ -static inline void -tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len; - - len = xprt->tcp_reclen - xprt->tcp_offset; - if (len > desc->count) - len = desc->count; - desc->count -= len; - desc->offset += len; - xprt->tcp_offset += len; - dprintk("RPC: discarded %Zu bytes\n", len); - tcp_check_recm(xprt); -} - -/* - * TCP record receive routine - * We first have to grab the record marker, then the XID, then the data. +/** + * xprt_update_rtt - update an RPC client's RTT state after receiving a reply + * @task: RPC request that recently completed + * */ -static int -tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, - unsigned int offset, size_t len) +void xprt_update_rtt(struct rpc_task *task) { - struct rpc_xprt *xprt = rd_desc->arg.data; - skb_reader_t desc = { - .skb = skb, - .offset = offset, - .count = len, - .csum = 0 - }; - - dprintk("RPC: tcp_data_recv\n"); - do { - /* Read in a new fragment marker if necessary */ - /* Can we ever really expect to get completely empty fragments? */ - if (xprt->tcp_flags & XPRT_COPY_RECM) { - tcp_read_fraghdr(xprt, &desc); - continue; - } - /* Read in the xid if necessary */ - if (xprt->tcp_flags & XPRT_COPY_XID) { - tcp_read_xid(xprt, &desc); - continue; - } - /* Read in the request data */ - if (xprt->tcp_flags & XPRT_COPY_DATA) { - tcp_read_request(xprt, &desc); - continue; - } - /* Skip over any trailing bytes on short reads */ - tcp_read_discard(xprt, &desc); - } while (desc.count); - dprintk("RPC: tcp_data_recv done\n"); - return len - desc.count; -} - -static void tcp_data_ready(struct sock *sk, int bytes) -{ - struct rpc_xprt *xprt; - read_descriptor_t rd_desc; - - read_lock(&sk->sk_callback_lock); - dprintk("RPC: tcp_data_ready...\n"); - if (!(xprt = xprt_from_sock(sk))) { - printk("RPC: tcp_data_ready socket info not found!\n"); - goto out; - } - if (xprt->shutdown) - goto out; - - /* We use rd_desc to pass struct xprt to tcp_data_recv */ - rd_desc.arg.data = xprt; - rd_desc.count = 65536; - tcp_read_sock(sk, &rd_desc, tcp_data_recv); -out: - read_unlock(&sk->sk_callback_lock); -} - -static void -tcp_state_change(struct sock *sk) -{ - struct rpc_xprt *xprt; + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_rtt *rtt = task->tk_client->cl_rtt; + unsigned timer = task->tk_msg.rpc_proc->p_timer; - read_lock(&sk->sk_callback_lock); - if (!(xprt = xprt_from_sock(sk))) - goto out; - dprintk("RPC: tcp_state_change client %p...\n", xprt); - dprintk("RPC: state %x conn %d dead %d zapped %d\n", - sk->sk_state, xprt_connected(xprt), - sock_flag(sk, SOCK_DEAD), - sock_flag(sk, SOCK_ZAPPED)); - - switch (sk->sk_state) { - case TCP_ESTABLISHED: - spin_lock_bh(&xprt->sock_lock); - if (!xprt_test_and_set_connected(xprt)) { - /* Reset TCP record info */ - xprt->tcp_offset = 0; - xprt->tcp_reclen = 0; - xprt->tcp_copied = 0; - xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID; - rpc_wake_up(&xprt->pending); - } - spin_unlock_bh(&xprt->sock_lock); - break; - case TCP_SYN_SENT: - case TCP_SYN_RECV: - break; - default: - xprt_disconnect(xprt); - break; + if (timer) { + if (req->rq_ntrans == 1) + rpc_update_rtt(rtt, timer, + (long)jiffies - req->rq_xtime); + rpc_set_timeo(rtt, timer, req->rq_ntrans - 1); } - out: - read_unlock(&sk->sk_callback_lock); } -/* - * Called when more output buffer space is available for this socket. - * We try not to wake our writers until they can make "significant" - * progress, otherwise we'll waste resources thrashing sock_sendmsg - * with a bunch of small requests. +/** + * xprt_complete_rqst - called when reply processing is complete + * @task: RPC request that recently completed + * @copied: actual number of bytes received from the transport + * + * Caller holds transport lock. */ -static void -xprt_write_space(struct sock *sk) +void xprt_complete_rqst(struct rpc_task *task, int copied) { - struct rpc_xprt *xprt; - struct socket *sock; - - read_lock(&sk->sk_callback_lock); - if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket)) - goto out; - if (xprt->shutdown) - goto out; - - /* Wait until we have enough socket memory */ - if (xprt->stream) { - /* from net/core/stream.c:sk_stream_write_space */ - if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk)) - goto out; - } else { - /* from net/core/sock.c:sock_def_write_space */ - if (!sock_writeable(sk)) - goto out; - } + struct rpc_rqst *req = task->tk_rqstp; - if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags)) - goto out; + dprintk("RPC: %5u xid %08x complete (%d bytes received)\n", + task->tk_pid, ntohl(req->rq_xid), copied); - spin_lock_bh(&xprt->sock_lock); - if (xprt->snd_task) - rpc_wake_up_task(xprt->snd_task); - spin_unlock_bh(&xprt->sock_lock); -out: - read_unlock(&sk->sk_callback_lock); + list_del_init(&req->rq_list); + req->rq_received = req->rq_private_buf.len = copied; + rpc_wake_up_task(task); } -/* - * RPC receive timeout handler. - */ -static void -xprt_timer(struct rpc_task *task) +static void xprt_timer(struct rpc_task *task) { - struct rpc_rqst *req = task->tk_rqstp; + struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; - spin_lock(&xprt->sock_lock); - if (req->rq_received) - goto out; + dprintk("RPC: %4d xprt_timer\n", task->tk_pid); - xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT); - __xprt_put_cong(xprt, req); - - dprintk("RPC: %4d xprt_timer (%s request)\n", - task->tk_pid, req ? "pending" : "backlogged"); - - task->tk_status = -ETIMEDOUT; -out: + spin_lock(&xprt->transport_lock); + if (!req->rq_received) { + if (xprt->ops->timer) + xprt->ops->timer(task); + task->tk_status = -ETIMEDOUT; + } task->tk_timeout = 0; rpc_wake_up_task(task); - spin_unlock(&xprt->sock_lock); + spin_unlock(&xprt->transport_lock); } -/* - * Place the actual RPC call. - * We have to copy the iovec because sendmsg fiddles with its contents. +/** + * xprt_prepare_transmit - reserve the transport before sending a request + * @task: RPC task about to send a request + * */ -int -xprt_prepare_transmit(struct rpc_task *task) +int xprt_prepare_transmit(struct rpc_task *task) { struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; @@ -1191,12 +690,12 @@ xprt_prepare_transmit(struct rpc_task *t if (xprt->shutdown) return -EIO; - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); if (req->rq_received && !req->rq_bytes_sent) { err = req->rq_received; goto out_unlock; } - if (!__xprt_lock_write(xprt, task)) { + if (!xprt->ops->reserve_xprt(task)) { err = -EAGAIN; goto out_unlock; } @@ -1206,39 +705,42 @@ xprt_prepare_transmit(struct rpc_task *t goto out_unlock; } out_unlock: - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); return err; } void -xprt_transmit(struct rpc_task *task) +xprt_abort_transmit(struct rpc_task *task) +{ + struct rpc_xprt *xprt = task->tk_xprt; + + xprt_release_write(xprt, task); +} + +/** + * xprt_transmit - send an RPC request on a transport + * @task: controlling RPC task + * + * We have to copy the iovec because sendmsg fiddles with its contents. + */ +void xprt_transmit(struct rpc_task *task) { - struct rpc_clnt *clnt = task->tk_client; struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; - int status, retry = 0; - + int status; dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen); - /* set up everything as needed. */ - /* Write the record marker */ - if (xprt->stream) { - u32 *marker = req->rq_svec[0].iov_base; - - *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker))); - } - smp_rmb(); if (!req->rq_received) { if (list_empty(&req->rq_list)) { - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); /* Update the softirq receive buffer */ memcpy(&req->rq_private_buf, &req->rq_rcv_buf, sizeof(req->rq_private_buf)); /* Add request to the receive list */ list_add_tail(&req->rq_list, &xprt->recv); - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); xprt_reset_majortimeo(req); /* Turn off autodisconnect */ del_singleshot_timer_sync(&xprt->timer); @@ -1246,40 +748,19 @@ xprt_transmit(struct rpc_task *task) } else if (!req->rq_bytes_sent) return; - /* Continue transmitting the packet/record. We must be careful - * to cope with writespace callbacks arriving _after_ we have - * called xprt_sendmsg(). - */ - while (1) { - req->rq_xtime = jiffies; - status = xprt_sendmsg(xprt, req); - - if (status < 0) - break; - - if (xprt->stream) { - req->rq_bytes_sent += status; - - /* If we've sent the entire packet, immediately - * reset the count of bytes sent. */ - if (req->rq_bytes_sent >= req->rq_slen) { - req->rq_bytes_sent = 0; - goto out_receive; - } - } else { - if (status >= req->rq_slen) - goto out_receive; - status = -EAGAIN; - break; - } - - dprintk("RPC: %4d xmit incomplete (%d left of %d)\n", - task->tk_pid, req->rq_slen - req->rq_bytes_sent, - req->rq_slen); - - status = -EAGAIN; - if (retry++ > 50) - break; + status = xprt->ops->send_request(task); + if (status == 0) { + dprintk("RPC: %4d xmit complete\n", task->tk_pid); + spin_lock_bh(&xprt->transport_lock); + xprt->ops->set_retrans_timeout(task); + /* Don't race with disconnect */ + if (!xprt_connected(xprt)) + task->tk_status = -ENOTCONN; + else if (!req->rq_received) + rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer); + xprt->ops->release_xprt(xprt, task); + spin_unlock_bh(&xprt->transport_lock); + return; } /* Note: at this point, task->tk_sleeping has not yet been set, @@ -1289,60 +770,19 @@ xprt_transmit(struct rpc_task *task) task->tk_status = status; switch (status) { - case -EAGAIN: - if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) { - /* Protect against races with xprt_write_space */ - spin_lock_bh(&xprt->sock_lock); - /* Don't race with disconnect */ - if (!xprt_connected(xprt)) - task->tk_status = -ENOTCONN; - else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) { - task->tk_timeout = req->rq_timeout; - rpc_sleep_on(&xprt->pending, task, NULL, NULL); - } - spin_unlock_bh(&xprt->sock_lock); - return; - } - /* Keep holding the socket if it is blocked */ - rpc_delay(task, HZ>>4); - return; case -ECONNREFUSED: - task->tk_timeout = RPC_REESTABLISH_TIMEOUT; rpc_sleep_on(&xprt->sending, task, NULL, NULL); + case -EAGAIN: case -ENOTCONN: return; default: - if (xprt->stream) - xprt_disconnect(xprt); + break; } xprt_release_write(xprt, task); return; - out_receive: - dprintk("RPC: %4d xmit complete\n", task->tk_pid); - /* Set the task's receive timeout value */ - spin_lock_bh(&xprt->sock_lock); - if (!xprt->nocong) { - int timer = task->tk_msg.rpc_proc->p_timer; - task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer); - task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries; - if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0) - task->tk_timeout = xprt->timeout.to_maxval; - } else - task->tk_timeout = req->rq_timeout; - /* Don't race with disconnect */ - if (!xprt_connected(xprt)) - task->tk_status = -ENOTCONN; - else if (!req->rq_received) - rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer); - __xprt_release_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); } -/* - * Reserve an RPC call slot. - */ -static inline void -do_xprt_reserve(struct rpc_task *task) +static inline void do_xprt_reserve(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; @@ -1362,22 +802,25 @@ do_xprt_reserve(struct rpc_task *task) rpc_sleep_on(&xprt->backlog, task, NULL, NULL); } -void -xprt_reserve(struct rpc_task *task) +/** + * xprt_reserve - allocate an RPC request slot + * @task: RPC task requesting a slot allocation + * + * If no more slots are available, place the task on the transport's + * backlog queue. + */ +void xprt_reserve(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; task->tk_status = -EIO; if (!xprt->shutdown) { - spin_lock(&xprt->xprt_lock); + spin_lock(&xprt->reserve_lock); do_xprt_reserve(task); - spin_unlock(&xprt->xprt_lock); + spin_unlock(&xprt->reserve_lock); } } -/* - * Allocate a 'unique' XID - */ static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt) { return xprt->xid++; @@ -1388,11 +831,7 @@ static inline void xprt_init_xid(struct get_random_bytes(&xprt->xid, sizeof(xprt->xid)); } -/* - * Initialize RPC request - */ -static void -xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt) +static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt) { struct rpc_rqst *req = task->tk_rqstp; @@ -1404,124 +843,97 @@ xprt_request_init(struct rpc_task *task, req, ntohl(req->rq_xid)); } -/* - * Release an RPC call slot +/** + * xprt_release - release an RPC request slot + * @task: task which is finished with the slot + * */ -void -xprt_release(struct rpc_task *task) +void xprt_release(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; struct rpc_rqst *req; if (!(req = task->tk_rqstp)) return; - spin_lock_bh(&xprt->sock_lock); - __xprt_release_write(xprt, task); - __xprt_put_cong(xprt, req); + spin_lock_bh(&xprt->transport_lock); + xprt->ops->release_xprt(xprt, task); + if (xprt->ops->release_request) + xprt->ops->release_request(task); if (!list_empty(&req->rq_list)) list_del(&req->rq_list); xprt->last_used = jiffies; if (list_empty(&xprt->recv) && !xprt->shutdown) - mod_timer(&xprt->timer, xprt->last_used + XPRT_IDLE_TIMEOUT); - spin_unlock_bh(&xprt->sock_lock); + mod_timer(&xprt->timer, + xprt->last_used + xprt->idle_timeout); + spin_unlock_bh(&xprt->transport_lock); task->tk_rqstp = NULL; memset(req, 0, sizeof(*req)); /* mark unused */ dprintk("RPC: %4d release request %p\n", task->tk_pid, req); - spin_lock(&xprt->xprt_lock); + spin_lock(&xprt->reserve_lock); list_add(&req->rq_list, &xprt->free); - xprt_clear_backlog(xprt); - spin_unlock(&xprt->xprt_lock); -} - -/* - * Set default timeout parameters - */ -static void -xprt_default_timeout(struct rpc_timeout *to, int proto) -{ - if (proto == IPPROTO_UDP) - xprt_set_timeout(to, 5, 5 * HZ); - else - xprt_set_timeout(to, 5, 60 * HZ); + rpc_wake_up_next(&xprt->backlog); + spin_unlock(&xprt->reserve_lock); } -/* - * Set constant timeout +/** + * xprt_set_timeout - set constant RPC timeout + * @to: RPC timeout parameters to set up + * @retr: number of retries + * @incr: amount of increase after each retry + * */ -void -xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr) +void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr) { to->to_initval = to->to_increment = incr; - to->to_maxval = incr * retr; + to->to_maxval = to->to_initval + (incr * retr); to->to_retries = retr; to->to_exponential = 0; } -unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; -unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; - -/* - * Initialize an RPC client - */ -static struct rpc_xprt * -xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to) +static struct rpc_xprt *xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to) { + int result; struct rpc_xprt *xprt; - unsigned int entries; - size_t slot_table_size; struct rpc_rqst *req; - dprintk("RPC: setting up %s transport...\n", - proto == IPPROTO_UDP? "UDP" : "TCP"); - - entries = (proto == IPPROTO_TCP)? - xprt_tcp_slot_table_entries : xprt_udp_slot_table_entries; - if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL) return ERR_PTR(-ENOMEM); memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */ - xprt->max_reqs = entries; - slot_table_size = entries * sizeof(xprt->slot[0]); - xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); - if (xprt->slot == NULL) { - kfree(xprt); - return ERR_PTR(-ENOMEM); - } - memset(xprt->slot, 0, slot_table_size); xprt->addr = *ap; - xprt->prot = proto; - xprt->stream = (proto == IPPROTO_TCP)? 1 : 0; - if (xprt->stream) { - xprt->cwnd = RPC_MAXCWND(xprt); - xprt->nocong = 1; - xprt->max_payload = (1U << 31) - 1; - } else { - xprt->cwnd = RPC_INITCWND; - xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); + + switch (proto) { + case IPPROTO_UDP: + result = xs_setup_udp(xprt, to); + break; + case IPPROTO_TCP: + result = xs_setup_tcp(xprt, to); + break; + default: + printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n", + proto); + result = -EIO; + break; } - spin_lock_init(&xprt->sock_lock); - spin_lock_init(&xprt->xprt_lock); - init_waitqueue_head(&xprt->cong_wait); + if (result) { + kfree(xprt); + return ERR_PTR(result); + } + + spin_lock_init(&xprt->transport_lock); + spin_lock_init(&xprt->reserve_lock); INIT_LIST_HEAD(&xprt->free); INIT_LIST_HEAD(&xprt->recv); - INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt); - INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt); + INIT_WORK(&xprt->task_cleanup, xprt_autoclose, xprt); init_timer(&xprt->timer); xprt->timer.function = xprt_init_autodisconnect; xprt->timer.data = (unsigned long) xprt; xprt->last_used = jiffies; - xprt->port = XPRT_MAX_RESVPORT; - - /* Set timeout parameters */ - if (to) { - xprt->timeout = *to; - } else - xprt_default_timeout(&xprt->timeout, xprt->prot); + xprt->cwnd = RPC_INITCWND; rpc_init_wait_queue(&xprt->pending, "xprt_pending"); rpc_init_wait_queue(&xprt->sending, "xprt_sending"); @@ -1529,139 +941,25 @@ xprt_setup(int proto, struct sockaddr_in rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog"); /* initialize free list */ - for (req = &xprt->slot[entries-1]; req >= &xprt->slot[0]; req--) + for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--) list_add(&req->rq_list, &xprt->free); xprt_init_xid(xprt); - /* Check whether we want to use a reserved port */ - xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; - dprintk("RPC: created transport %p with %u slots\n", xprt, xprt->max_reqs); return xprt; } -/* - * Bind to a reserved port - */ -static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock) -{ - struct sockaddr_in myaddr = { - .sin_family = AF_INET, - }; - int err, port; - - /* Were we already bound to a given port? Try to reuse it */ - port = xprt->port; - do { - myaddr.sin_port = htons(port); - err = sock->ops->bind(sock, (struct sockaddr *) &myaddr, - sizeof(myaddr)); - if (err == 0) { - xprt->port = port; - return 0; - } - if (--port == 0) - port = XPRT_MAX_RESVPORT; - } while (err == -EADDRINUSE && port != xprt->port); - - printk("RPC: Can't bind to reserved port (%d).\n", -err); - return err; -} - -static void -xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock) -{ - struct sock *sk = sock->sk; - - if (xprt->inet) - return; - - write_lock_bh(&sk->sk_callback_lock); - sk->sk_user_data = xprt; - xprt->old_data_ready = sk->sk_data_ready; - xprt->old_state_change = sk->sk_state_change; - xprt->old_write_space = sk->sk_write_space; - if (xprt->prot == IPPROTO_UDP) { - sk->sk_data_ready = udp_data_ready; - sk->sk_no_check = UDP_CSUM_NORCV; - xprt_set_connected(xprt); - } else { - tcp_sk(sk)->nonagle = 1; /* disable Nagle's algorithm */ - sk->sk_data_ready = tcp_data_ready; - sk->sk_state_change = tcp_state_change; - xprt_clear_connected(xprt); - } - sk->sk_write_space = xprt_write_space; - - /* Reset to new socket */ - xprt->sock = sock; - xprt->inet = sk; - write_unlock_bh(&sk->sk_callback_lock); - - return; -} - -/* - * Set socket buffer length - */ -void -xprt_sock_setbufsize(struct rpc_xprt *xprt) -{ - struct sock *sk = xprt->inet; - - if (xprt->stream) - return; - if (xprt->rcvsize) { - sk->sk_userlocks |= SOCK_RCVBUF_LOCK; - sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2; - } - if (xprt->sndsize) { - sk->sk_userlocks |= SOCK_SNDBUF_LOCK; - sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2; - sk->sk_write_space(sk); - } -} - -/* - * Datastream sockets are created here, but xprt_connect will create - * and connect stream sockets. - */ -static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport) -{ - struct socket *sock; - int type, err; - - dprintk("RPC: xprt_create_socket(%s %d)\n", - (proto == IPPROTO_UDP)? "udp" : "tcp", proto); - - type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM; - - if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) { - printk("RPC: can't create socket (%d).\n", -err); - return NULL; - } - - /* If the caller has the capability, bind to a reserved port */ - if (resvport && xprt_bindresvport(xprt, sock) < 0) { - printk("RPC: can't bind to reserved port.\n"); - goto failed; - } - - return sock; - -failed: - sock_release(sock); - return NULL; -} - -/* - * Create an RPC client transport given the protocol and peer address. +/** + * xprt_create_proto - create an RPC client transport + * @proto: requested transport protocol + * @sap: remote peer's address + * @to: timeout parameters for new transport + * */ -struct rpc_xprt * -xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to) +struct rpc_xprt *xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to) { struct rpc_xprt *xprt; @@ -1673,46 +971,26 @@ xprt_create_proto(int proto, struct sock return xprt; } -/* - * Prepare for transport shutdown. - */ -static void -xprt_shutdown(struct rpc_xprt *xprt) +static void xprt_shutdown(struct rpc_xprt *xprt) { xprt->shutdown = 1; rpc_wake_up(&xprt->sending); rpc_wake_up(&xprt->resend); - rpc_wake_up(&xprt->pending); + xprt_wake_pending_tasks(xprt, -EIO); rpc_wake_up(&xprt->backlog); - wake_up(&xprt->cong_wait); del_timer_sync(&xprt->timer); - - /* synchronously wait for connect worker to finish */ - cancel_delayed_work(&xprt->sock_connect); - flush_scheduled_work(); -} - -/* - * Clear the xprt backlog queue - */ -static int -xprt_clear_backlog(struct rpc_xprt *xprt) { - rpc_wake_up_next(&xprt->backlog); - wake_up(&xprt->cong_wait); - return 1; } -/* - * Destroy an RPC transport, killing off all requests. +/** + * xprt_destroy - destroy an RPC transport, killing off all requests. + * @xprt: transport to destroy + * */ -int -xprt_destroy(struct rpc_xprt *xprt) +int xprt_destroy(struct rpc_xprt *xprt) { dprintk("RPC: destroying transport %p\n", xprt); xprt_shutdown(xprt); - xprt_disconnect(xprt); - xprt_close(xprt); - kfree(xprt->slot); + xprt->ops->destroy(xprt); kfree(xprt); return 0; Index: linux-2.6/net/sunrpc/xprtsock.c =================================================================== --- /dev/null +++ linux-2.6/net/sunrpc/xprtsock.c @@ -0,0 +1,1252 @@ +/* + * linux/net/sunrpc/xprtsock.c + * + * Client-side transport implementation for sockets. + * + * TCP callback races fixes (C) 1998 Red Hat Software + * TCP send fixes (C) 1998 Red Hat Software + * TCP NFS related read + write fixes + * (C) 1999 Dave Airlie, University of Limerick, Ireland + * + * Rewrite of larges part of the code in order to stabilize TCP stuff. + * Fix behaviour when socket buffer is full. + * (C) 1999 Trond Myklebust + * + * IP socket transport implementation, (C) 2005 Chuck Lever + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +/* + * How many times to try sending a request on a socket before waiting + * for the socket buffer to clear. + */ +#define XS_SENDMSG_RETRY (10U) + +/* + * Time out for an RPC UDP socket connect. UDP socket connects are + * synchronous, but we set a timeout anyway in case of resource + * exhaustion on the local host. + */ +#define XS_UDP_CONN_TO (5U * HZ) + +/* + * Wait duration for an RPC TCP connection to be established. Solaris + * NFS over TCP uses 60 seconds, for example, which is in line with how + * long a server takes to reboot. + */ +#define XS_TCP_CONN_TO (60U * HZ) + +/* + * Wait duration for a reply from the RPC portmapper. + */ +#define XS_BIND_TO (60U * HZ) + +/* + * Delay if a UDP socket connect error occurs. This is most likely some + * kind of resource problem on the local host. + */ +#define XS_UDP_REEST_TO (2U * HZ) + +/* + * The reestablish timeout allows clients to delay for a bit before attempting + * to reconnect to a server that just dropped our connection. + * + * We implement an exponential backoff when trying to reestablish a TCP + * transport connection with the server. Some servers like to drop a TCP + * connection when they are overworked, so we start with a short timeout and + * increase over time if the server is down or not responding. + */ +#define XS_TCP_INIT_REEST_TO (3U * HZ) +#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) + +/* + * TCP idle timeout; client drops the transport socket if it is idle + * for this long. Note that we also timeout UDP sockets to prevent + * holding port numbers when there is no RPC traffic. + */ +#define XS_IDLE_DISC_TO (5U * 60 * HZ) + +#ifdef RPC_DEBUG +# undef RPC_DEBUG_DATA +# define RPCDBG_FACILITY RPCDBG_TRANS +#endif + +#ifdef RPC_DEBUG_DATA +static void xs_pktdump(char *msg, u32 *packet, unsigned int count) +{ + u8 *buf = (u8 *) packet; + int j; + + dprintk("RPC: %s\n", msg); + for (j = 0; j < count && j < 128; j += 4) { + if (!(j & 31)) { + if (j) + dprintk("\n"); + dprintk("0x%04x ", j); + } + dprintk("%02x%02x%02x%02x ", + buf[j], buf[j+1], buf[j+2], buf[j+3]); + } + dprintk("\n"); +} +#else +static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) +{ + /* NOP */ +} +#endif + +#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) + +static inline int xs_send_head(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, unsigned int len) +{ + struct kvec iov = { + .iov_base = xdr->head[0].iov_base + base, + .iov_len = len - base, + }; + struct msghdr msg = { + .msg_name = addr, + .msg_namelen = addrlen, + .msg_flags = XS_SENDMSG_FLAGS, + }; + + if (xdr->len > len) + msg.msg_flags |= MSG_MORE; + + if (likely(iov.iov_len)) + return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); + return kernel_sendmsg(sock, &msg, NULL, 0, 0); +} + +static int xs_send_tail(struct socket *sock, struct xdr_buf *xdr, unsigned int base, unsigned int len) +{ + struct kvec iov = { + .iov_base = xdr->tail[0].iov_base + base, + .iov_len = len - base, + }; + struct msghdr msg = { + .msg_flags = XS_SENDMSG_FLAGS, + }; + + return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); +} + +/** + * xs_sendpages - write pages directly to a socket + * @sock: socket to send on + * @addr: UDP only -- address of destination + * @addrlen: UDP only -- length of destination address + * @xdr: buffer containing this request + * @base: starting position in the buffer + * + */ +static inline int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) +{ + struct page **ppage = xdr->pages; + unsigned int len, pglen = xdr->page_len; + int err, ret = 0; + ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int); + + if (unlikely(!sock)) + return -ENOTCONN; + + clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); + + len = xdr->head[0].iov_len; + if (base < len || (addr != NULL && base == 0)) { + err = xs_send_head(sock, addr, addrlen, xdr, base, len); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + if (err != (len - base)) + goto out; + base = 0; + } else + base -= len; + + if (unlikely(pglen == 0)) + goto copy_tail; + if (unlikely(base >= pglen)) { + base -= pglen; + goto copy_tail; + } + if (base || xdr->page_base) { + pglen -= base; + base += xdr->page_base; + ppage += base >> PAGE_CACHE_SHIFT; + base &= ~PAGE_CACHE_MASK; + } + + sendpage = sock->ops->sendpage ? : sock_no_sendpage; + do { + int flags = XS_SENDMSG_FLAGS; + + len = PAGE_CACHE_SIZE; + if (base) + len -= base; + if (pglen < len) + len = pglen; + + if (pglen != len || xdr->tail[0].iov_len != 0) + flags |= MSG_MORE; + + /* Hmm... We might be dealing with highmem pages */ + if (PageHighMem(*ppage)) + sendpage = sock_no_sendpage; + err = sendpage(sock, *ppage, base, len, flags); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + if (err != len) + goto out; + base = 0; + ppage++; + } while ((pglen -= len) != 0); +copy_tail: + len = xdr->tail[0].iov_len; + if (base < len) { + err = xs_send_tail(sock, xdr, base, len); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + } +out: + return ret; +} + +/** + * xs_nospace - place task on wait queue if transmit was incomplete + * @task: task to put to sleep + * + */ +static void xs_nospace(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + + dprintk("RPC: %4d xmit incomplete (%u left of %u)\n", + task->tk_pid, req->rq_slen - req->rq_bytes_sent, + req->rq_slen); + + if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) { + /* Protect against races with write_space */ + spin_lock_bh(&xprt->transport_lock); + + /* Don't race with disconnect */ + if (!xprt_connected(xprt)) + task->tk_status = -ENOTCONN; + else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) + xprt_wait_for_buffer_space(task); + + spin_unlock_bh(&xprt->transport_lock); + } else + /* Keep holding the socket if it is blocked */ + rpc_delay(task, HZ>>4); +} + +/** + * xs_udp_send_request - write an RPC request to a UDP socket + * @task: address of RPC task that manages the state of an RPC request + * + * Return values: + * 0: The request has been sent + * EAGAIN: The socket was blocked, please call again later to + * complete the request + * ENOTCONN: Caller needs to invoke connect logic then call again + * other: Some other error occured, the request was not sent + */ +static int xs_udp_send_request(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + struct xdr_buf *xdr = &req->rq_snd_buf; + int status; + + xs_pktdump("packet data:", + req->rq_svec->iov_base, + req->rq_svec->iov_len); + + req->rq_xtime = jiffies; + status = xs_sendpages(xprt->sock, (struct sockaddr *) &xprt->addr, + sizeof(xprt->addr), xdr, req->rq_bytes_sent); + + dprintk("RPC: xs_udp_send_request(%u) = %d\n", + xdr->len - req->rq_bytes_sent, status); + + if (likely(status >= (int) req->rq_slen)) + return 0; + + /* Still some bytes left; set up for a retry later. */ + if (status > 0) + status = -EAGAIN; + + switch (status) { + case -ENETUNREACH: + case -EPIPE: + case -ECONNREFUSED: + /* When the server has died, an ICMP port unreachable message + * prompts ECONNREFUSED. */ + break; + case -EAGAIN: + xs_nospace(task); + break; + default: + dprintk("RPC: sendmsg returned unrecognized error %d\n", + -status); + break; + } + + return status; +} + +static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf) +{ + u32 reclen = buf->len - sizeof(rpc_fraghdr); + rpc_fraghdr *base = buf->head[0].iov_base; + *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen); +} + +/** + * xs_tcp_send_request - write an RPC request to a TCP socket + * @task: address of RPC task that manages the state of an RPC request + * + * Return values: + * 0: The request has been sent + * EAGAIN: The socket was blocked, please call again later to + * complete the request + * ENOTCONN: Caller needs to invoke connect logic then call again + * other: Some other error occured, the request was not sent + * + * XXX: In the case of soft timeouts, should we eventually give up + * if sendmsg is not able to make progress? + */ +static int xs_tcp_send_request(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + struct xdr_buf *xdr = &req->rq_snd_buf; + int status, retry = 0; + + xs_encode_tcp_record_marker(&req->rq_snd_buf); + + xs_pktdump("packet data:", + req->rq_svec->iov_base, + req->rq_svec->iov_len); + + /* Continue transmitting the packet/record. We must be careful + * to cope with writespace callbacks arriving _after_ we have + * called sendmsg(). */ + while (1) { + req->rq_xtime = jiffies; + status = xs_sendpages(xprt->sock, NULL, 0, xdr, + req->rq_bytes_sent); + + dprintk("RPC: xs_tcp_send_request(%u) = %d\n", + xdr->len - req->rq_bytes_sent, status); + + if (unlikely(status < 0)) + break; + + /* If we've sent the entire packet, immediately + * reset the count of bytes sent. */ + req->rq_bytes_sent += status; + if (likely(req->rq_bytes_sent >= req->rq_slen)) { + req->rq_bytes_sent = 0; + return 0; + } + + status = -EAGAIN; + if (retry++ > XS_SENDMSG_RETRY) + break; + } + + switch (status) { + case -EAGAIN: + xs_nospace(task); + break; + case -ECONNREFUSED: + case -ECONNRESET: + case -ENOTCONN: + case -EPIPE: + status = -ENOTCONN; + break; + default: + dprintk("RPC: sendmsg returned unrecognized error %d\n", + -status); + xprt_disconnect(xprt); + break; + } + + return status; +} + +/** + * xs_close - close a socket + * @xprt: transport + * + * This is used when all requests are complete; ie, no DRC state remains + * on the server we want to save. + */ +static void xs_close(struct rpc_xprt *xprt) +{ + struct socket *sock = xprt->sock; + struct sock *sk = xprt->inet; + + if (!sk) + return; + + dprintk("RPC: xs_close xprt %p\n", xprt); + + write_lock_bh(&sk->sk_callback_lock); + xprt->inet = NULL; + xprt->sock = NULL; + + sk->sk_user_data = NULL; + sk->sk_data_ready = xprt->old_data_ready; + sk->sk_state_change = xprt->old_state_change; + sk->sk_write_space = xprt->old_write_space; + write_unlock_bh(&sk->sk_callback_lock); + + sk->sk_no_check = 0; + + sock_release(sock); +} + +/** + * xs_destroy - prepare to shutdown a transport + * @xprt: doomed transport + * + */ +static void xs_destroy(struct rpc_xprt *xprt) +{ + dprintk("RPC: xs_destroy xprt %p\n", xprt); + + cancel_delayed_work(&xprt->connect_worker); + flush_scheduled_work(); + + xprt_disconnect(xprt); + xs_close(xprt); + kfree(xprt->slot); +} + +static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) +{ + return (struct rpc_xprt *) sk->sk_user_data; +} + +/** + * xs_udp_data_ready - "data ready" callback for UDP sockets + * @sk: socket with data to read + * @len: how much data to read + * + */ +static void xs_udp_data_ready(struct sock *sk, int len) +{ + struct rpc_task *task; + struct rpc_xprt *xprt; + struct rpc_rqst *rovr; + struct sk_buff *skb; + int err, repsize, copied; + u32 _xid, *xp; + + read_lock(&sk->sk_callback_lock); + dprintk("RPC: xs_udp_data_ready...\n"); + if (!(xprt = xprt_from_sock(sk))) + goto out; + + if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) + goto out; + + if (xprt->shutdown) + goto dropit; + + repsize = skb->len - sizeof(struct udphdr); + if (repsize < 4) { + dprintk("RPC: impossible RPC reply size %d!\n", repsize); + goto dropit; + } + + /* Copy the XID from the skb... */ + xp = skb_header_pointer(skb, sizeof(struct udphdr), + sizeof(_xid), &_xid); + if (xp == NULL) + goto dropit; + + /* Look up and lock the request corresponding to the given XID */ + spin_lock(&xprt->transport_lock); + rovr = xprt_lookup_rqst(xprt, *xp); + if (!rovr) + goto out_unlock; + task = rovr->rq_task; + + if ((copied = rovr->rq_private_buf.buflen) > repsize) + copied = repsize; + + /* Suck it into the iovec, verify checksum if not done by hw. */ + if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) + goto out_unlock; + + /* Something worked... */ + dst_confirm(skb->dst); + + xprt_adjust_cwnd(task, copied); + xprt_update_rtt(task); + xprt_complete_rqst(task, copied); + + out_unlock: + spin_unlock(&xprt->transport_lock); + dropit: + skb_free_datagram(sk, skb); + out: + read_unlock(&sk->sk_callback_lock); +} + +static inline size_t xs_tcp_copy_data(skb_reader_t *desc, void *p, size_t len) +{ + if (len > desc->count) + len = desc->count; + if (skb_copy_bits(desc->skb, desc->offset, p, len)) { + dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n", + len, desc->count); + return 0; + } + desc->offset += len; + desc->count -= len; + dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n", + len, desc->count); + return len; +} + +static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len, used; + char *p; + + p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset; + len = sizeof(xprt->tcp_recm) - xprt->tcp_offset; + used = xs_tcp_copy_data(desc, p, len); + xprt->tcp_offset += used; + if (used != len) + return; + + xprt->tcp_reclen = ntohl(xprt->tcp_recm); + if (xprt->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) + xprt->tcp_flags |= XPRT_LAST_FRAG; + else + xprt->tcp_flags &= ~XPRT_LAST_FRAG; + xprt->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; + + xprt->tcp_flags &= ~XPRT_COPY_RECM; + xprt->tcp_offset = 0; + + /* Sanity check of the record length */ + if (unlikely(xprt->tcp_reclen < 4)) { + dprintk("RPC: invalid TCP record fragment length\n"); + xprt_disconnect(xprt); + return; + } + dprintk("RPC: reading TCP record fragment of length %d\n", + xprt->tcp_reclen); +} + +static void xs_tcp_check_recm(struct rpc_xprt *xprt) +{ + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags); + if (xprt->tcp_offset == xprt->tcp_reclen) { + xprt->tcp_flags |= XPRT_COPY_RECM; + xprt->tcp_offset = 0; + if (xprt->tcp_flags & XPRT_LAST_FRAG) { + xprt->tcp_flags &= ~XPRT_COPY_DATA; + xprt->tcp_flags |= XPRT_COPY_XID; + xprt->tcp_copied = 0; + } + } +} + +static inline void xs_tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len, used; + char *p; + + len = sizeof(xprt->tcp_xid) - xprt->tcp_offset; + dprintk("RPC: reading XID (%Zu bytes)\n", len); + p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset; + used = xs_tcp_copy_data(desc, p, len); + xprt->tcp_offset += used; + if (used != len) + return; + xprt->tcp_flags &= ~XPRT_COPY_XID; + xprt->tcp_flags |= XPRT_COPY_DATA; + xprt->tcp_copied = 4; + dprintk("RPC: reading reply for XID %08x\n", + ntohl(xprt->tcp_xid)); + xs_tcp_check_recm(xprt); +} + +static inline void xs_tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + struct rpc_rqst *req; + struct xdr_buf *rcvbuf; + size_t len; + ssize_t r; + + /* Find and lock the request corresponding to this xid */ + spin_lock(&xprt->transport_lock); + req = xprt_lookup_rqst(xprt, xprt->tcp_xid); + if (!req) { + xprt->tcp_flags &= ~XPRT_COPY_DATA; + dprintk("RPC: XID %08x request not found!\n", + ntohl(xprt->tcp_xid)); + spin_unlock(&xprt->transport_lock); + return; + } + + rcvbuf = &req->rq_private_buf; + len = desc->count; + if (len > xprt->tcp_reclen - xprt->tcp_offset) { + skb_reader_t my_desc; + + len = xprt->tcp_reclen - xprt->tcp_offset; + memcpy(&my_desc, desc, sizeof(my_desc)); + my_desc.count = len; + r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, + &my_desc, xs_tcp_copy_data); + desc->count -= r; + desc->offset += r; + } else + r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, + desc, xs_tcp_copy_data); + + if (r > 0) { + xprt->tcp_copied += r; + xprt->tcp_offset += r; + } + if (r != len) { + /* Error when copying to the receive buffer, + * usually because we weren't able to allocate + * additional buffer pages. All we can do now + * is turn off XPRT_COPY_DATA, so the request + * will not receive any additional updates, + * and time out. + * Any remaining data from this record will + * be discarded. + */ + xprt->tcp_flags &= ~XPRT_COPY_DATA; + dprintk("RPC: XID %08x truncated request\n", + ntohl(xprt->tcp_xid)); + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); + goto out; + } + + dprintk("RPC: XID %08x read %Zd bytes\n", + ntohl(xprt->tcp_xid), r); + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); + + if (xprt->tcp_copied == req->rq_private_buf.buflen) + xprt->tcp_flags &= ~XPRT_COPY_DATA; + else if (xprt->tcp_offset == xprt->tcp_reclen) { + if (xprt->tcp_flags & XPRT_LAST_FRAG) + xprt->tcp_flags &= ~XPRT_COPY_DATA; + } + +out: + if (!(xprt->tcp_flags & XPRT_COPY_DATA)) + xprt_complete_rqst(req->rq_task, xprt->tcp_copied); + spin_unlock(&xprt->transport_lock); + xs_tcp_check_recm(xprt); +} + +static inline void xs_tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len; + + len = xprt->tcp_reclen - xprt->tcp_offset; + if (len > desc->count) + len = desc->count; + desc->count -= len; + desc->offset += len; + xprt->tcp_offset += len; + dprintk("RPC: discarded %Zu bytes\n", len); + xs_tcp_check_recm(xprt); +} + +static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) +{ + struct rpc_xprt *xprt = rd_desc->arg.data; + skb_reader_t desc = { + .skb = skb, + .offset = offset, + .count = len, + .csum = 0 + }; + + dprintk("RPC: xs_tcp_data_recv started\n"); + do { + /* Read in a new fragment marker if necessary */ + /* Can we ever really expect to get completely empty fragments? */ + if (xprt->tcp_flags & XPRT_COPY_RECM) { + xs_tcp_read_fraghdr(xprt, &desc); + continue; + } + /* Read in the xid if necessary */ + if (xprt->tcp_flags & XPRT_COPY_XID) { + xs_tcp_read_xid(xprt, &desc); + continue; + } + /* Read in the request data */ + if (xprt->tcp_flags & XPRT_COPY_DATA) { + xs_tcp_read_request(xprt, &desc); + continue; + } + /* Skip over any trailing bytes on short reads */ + xs_tcp_read_discard(xprt, &desc); + } while (desc.count); + dprintk("RPC: xs_tcp_data_recv done\n"); + return len - desc.count; +} + +/** + * xs_tcp_data_ready - "data ready" callback for TCP sockets + * @sk: socket with data to read + * @bytes: how much data to read + * + */ +static void xs_tcp_data_ready(struct sock *sk, int bytes) +{ + struct rpc_xprt *xprt; + read_descriptor_t rd_desc; + + read_lock(&sk->sk_callback_lock); + dprintk("RPC: xs_tcp_data_ready...\n"); + if (!(xprt = xprt_from_sock(sk))) + goto out; + if (xprt->shutdown) + goto out; + + /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ + rd_desc.arg.data = xprt; + rd_desc.count = 65536; + tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); +out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_tcp_state_change - callback to handle TCP socket state changes + * @sk: socket whose state has changed + * + */ +static void xs_tcp_state_change(struct sock *sk) +{ + struct rpc_xprt *xprt; + + read_lock(&sk->sk_callback_lock); + if (!(xprt = xprt_from_sock(sk))) + goto out; + dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); + dprintk("RPC: state %x conn %d dead %d zapped %d\n", + sk->sk_state, xprt_connected(xprt), + sock_flag(sk, SOCK_DEAD), + sock_flag(sk, SOCK_ZAPPED)); + + switch (sk->sk_state) { + case TCP_ESTABLISHED: + spin_lock_bh(&xprt->transport_lock); + if (!xprt_test_and_set_connected(xprt)) { + /* Reset TCP record info */ + xprt->tcp_offset = 0; + xprt->tcp_reclen = 0; + xprt->tcp_copied = 0; + xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID; + xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; + xprt_wake_pending_tasks(xprt, 0); + } + spin_unlock_bh(&xprt->transport_lock); + break; + case TCP_SYN_SENT: + case TCP_SYN_RECV: + break; + default: + xprt_disconnect(xprt); + break; + } + out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_udp_write_space - callback invoked when socket buffer space + * becomes available + * @sk: socket whose state has changed + * + * Called when more output buffer space is available for this socket. + * We try not to wake our writers until they can make "significant" + * progress, otherwise we'll waste resources thrashing kernel_sendmsg + * with a bunch of small requests. + */ +static void xs_udp_write_space(struct sock *sk) +{ + read_lock(&sk->sk_callback_lock); + + /* from net/core/sock.c:sock_def_write_space */ + if (sock_writeable(sk)) { + struct socket *sock; + struct rpc_xprt *xprt; + + if (unlikely(!(sock = sk->sk_socket))) + goto out; + if (unlikely(!(xprt = xprt_from_sock(sk)))) + goto out; + if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + goto out; + + xprt_write_space(xprt); + } + + out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_tcp_write_space - callback invoked when socket buffer space + * becomes available + * @sk: socket whose state has changed + * + * Called when more output buffer space is available for this socket. + * We try not to wake our writers until they can make "significant" + * progress, otherwise we'll waste resources thrashing kernel_sendmsg + * with a bunch of small requests. + */ +static void xs_tcp_write_space(struct sock *sk) +{ + read_lock(&sk->sk_callback_lock); + + /* from net/core/stream.c:sk_stream_write_space */ + if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { + struct socket *sock; + struct rpc_xprt *xprt; + + if (unlikely(!(sock = sk->sk_socket))) + goto out; + if (unlikely(!(xprt = xprt_from_sock(sk)))) + goto out; + if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + goto out; + + xprt_write_space(xprt); + } + + out: + read_unlock(&sk->sk_callback_lock); +} + +static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) +{ + struct sock *sk = xprt->inet; + + if (xprt->rcvsize) { + sk->sk_userlocks |= SOCK_RCVBUF_LOCK; + sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2; + } + if (xprt->sndsize) { + sk->sk_userlocks |= SOCK_SNDBUF_LOCK; + sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2; + sk->sk_write_space(sk); + } +} + +/** + * xs_udp_set_buffer_size - set send and receive limits + * @xprt: generic transport + * @sndsize: requested size of send buffer, in bytes + * @rcvsize: requested size of receive buffer, in bytes + * + * Set socket send and receive buffer size limits. + */ +static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) +{ + xprt->sndsize = 0; + if (sndsize) + xprt->sndsize = sndsize + 1024; + xprt->rcvsize = 0; + if (rcvsize) + xprt->rcvsize = rcvsize + 1024; + + xs_udp_do_set_buffer_size(xprt); +} + +/** + * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport + * @task: task that timed out + * + * Adjust the congestion window after a retransmit timeout has occurred. + */ +static void xs_udp_timer(struct rpc_task *task) +{ + xprt_adjust_cwnd(task, -ETIMEDOUT); +} + +static int xs_bindresvport(struct rpc_xprt *xprt, struct socket *sock) +{ + struct sockaddr_in myaddr = { + .sin_family = AF_INET, + }; + int err; + unsigned short port = xprt->port; + + do { + myaddr.sin_port = htons(port); + err = sock->ops->bind(sock, (struct sockaddr *) &myaddr, + sizeof(myaddr)); + if (err == 0) { + xprt->port = port; + dprintk("RPC: xs_bindresvport bound to port %u\n", + port); + return 0; + } + if (port <= xprt_min_resvport) + port = xprt_max_resvport; + else + port--; + } while (err == -EADDRINUSE && port != xprt->port); + + dprintk("RPC: can't bind to reserved port (%d).\n", -err); + return err; +} + +/** + * xs_udp_connect_worker - set up a UDP socket + * @args: RPC transport to connect + * + * Invoked by a work queue tasklet. + */ +static void xs_udp_connect_worker(void *args) +{ + struct rpc_xprt *xprt = (struct rpc_xprt *) args; + struct socket *sock = xprt->sock; + int err, status = -EIO; + + if (xprt->shutdown || xprt->addr.sin_port == 0) + goto out; + + dprintk("RPC: xs_udp_connect_worker for xprt %p\n", xprt); + + /* Start by resetting any existing state */ + xs_close(xprt); + + if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { + dprintk("RPC: can't create UDP transport socket (%d).\n", -err); + goto out; + } + + if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { + sock_release(sock); + goto out; + } + + if (!xprt->inet) { + struct sock *sk = sock->sk; + + write_lock_bh(&sk->sk_callback_lock); + + sk->sk_user_data = xprt; + xprt->old_data_ready = sk->sk_data_ready; + xprt->old_state_change = sk->sk_state_change; + xprt->old_write_space = sk->sk_write_space; + sk->sk_data_ready = xs_udp_data_ready; + sk->sk_write_space = xs_udp_write_space; + sk->sk_no_check = UDP_CSUM_NORCV; + + xprt_set_connected(xprt); + + /* Reset to new socket */ + xprt->sock = sock; + xprt->inet = sk; + + write_unlock_bh(&sk->sk_callback_lock); + } + xs_udp_do_set_buffer_size(xprt); + status = 0; +out: + xprt_wake_pending_tasks(xprt, status); + xprt_clear_connecting(xprt); +} + +/* + * We need to preserve the port number so the reply cache on the server can + * find our cached RPC replies when we get around to reconnecting. + */ +static void xs_tcp_reuse_connection(struct rpc_xprt *xprt) +{ + int result; + struct socket *sock = xprt->sock; + struct sockaddr any; + + dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); + + /* + * Disconnect the transport socket by doing a connect operation + * with AF_UNSPEC. This should return immediately... + */ + memset(&any, 0, sizeof(any)); + any.sa_family = AF_UNSPEC; + result = sock->ops->connect(sock, &any, sizeof(any), 0); + if (result) + dprintk("RPC: AF_UNSPEC connect return code %d\n", + result); +} + +/** + * xs_tcp_connect_worker - connect a TCP socket to a remote endpoint + * @args: RPC transport to connect + * + * Invoked by a work queue tasklet. + */ +static void xs_tcp_connect_worker(void *args) +{ + struct rpc_xprt *xprt = (struct rpc_xprt *)args; + struct socket *sock = xprt->sock; + int err, status = -EIO; + + if (xprt->shutdown || xprt->addr.sin_port == 0) + goto out; + + dprintk("RPC: xs_tcp_connect_worker for xprt %p\n", xprt); + + if (!xprt->sock) { + /* start from scratch */ + if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { + dprintk("RPC: can't create TCP transport socket (%d).\n", -err); + goto out; + } + + if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { + sock_release(sock); + goto out; + } + } else + /* "close" the socket, preserving the local port */ + xs_tcp_reuse_connection(xprt); + + if (!xprt->inet) { + struct sock *sk = sock->sk; + + write_lock_bh(&sk->sk_callback_lock); + + sk->sk_user_data = xprt; + xprt->old_data_ready = sk->sk_data_ready; + xprt->old_state_change = sk->sk_state_change; + xprt->old_write_space = sk->sk_write_space; + sk->sk_data_ready = xs_tcp_data_ready; + sk->sk_state_change = xs_tcp_state_change; + sk->sk_write_space = xs_tcp_write_space; + + /* socket options */ + sk->sk_userlocks |= SOCK_BINDPORT_LOCK; + sock_reset_flag(sk, SOCK_LINGER); + tcp_sk(sk)->linger2 = 0; + tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; + + xprt_clear_connected(xprt); + + /* Reset to new socket */ + xprt->sock = sock; + xprt->inet = sk; + + write_unlock_bh(&sk->sk_callback_lock); + } + + /* Tell the socket layer to start connecting... */ + status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr, + sizeof(xprt->addr), O_NONBLOCK); + dprintk("RPC: %p connect status %d connected %d sock state %d\n", + xprt, -status, xprt_connected(xprt), sock->sk->sk_state); + if (status < 0) { + switch (status) { + case -EINPROGRESS: + case -EALREADY: + goto out_clear; + case -ECONNREFUSED: + case -ECONNRESET: + /* retry with existing socket, after a delay */ + break; + default: + /* get rid of existing socket, and retry */ + xs_close(xprt); + break; + } + } +out: + xprt_wake_pending_tasks(xprt, status); +out_clear: + xprt_clear_connecting(xprt); +} + +/** + * xs_connect - connect a socket to a remote endpoint + * @task: address of RPC task that manages state of connect request + * + * TCP: If the remote end dropped the connection, delay reconnecting. + * + * UDP socket connects are synchronous, but we use a work queue anyway + * to guarantee that even unprivileged user processes can set up a + * socket on a privileged port. + * + * If a UDP socket connect fails, the delay behavior here prevents + * retry floods (hard mounts). + */ +static void xs_connect(struct rpc_task *task) +{ + struct rpc_xprt *xprt = task->tk_xprt; + + if (xprt_test_and_set_connecting(xprt)) + return; + + if (xprt->sock != NULL) { + dprintk("RPC: xs_connect delayed xprt %p for %lu seconds\n", + xprt, xprt->reestablish_timeout / HZ); + schedule_delayed_work(&xprt->connect_worker, + xprt->reestablish_timeout); + xprt->reestablish_timeout <<= 1; + if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) + xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; + } else { + dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); + schedule_work(&xprt->connect_worker); + + /* flush_scheduled_work can sleep... */ + if (!RPC_IS_ASYNC(task)) + flush_scheduled_work(); + } +} + +static struct rpc_xprt_ops xs_udp_ops = { + .set_buffer_size = xs_udp_set_buffer_size, + .reserve_xprt = xprt_reserve_xprt_cong, + .release_xprt = xprt_release_xprt_cong, + .connect = xs_connect, + .send_request = xs_udp_send_request, + .set_retrans_timeout = xprt_set_retrans_timeout_rtt, + .timer = xs_udp_timer, + .release_request = xprt_release_rqst_cong, + .close = xs_close, + .destroy = xs_destroy, +}; + +static struct rpc_xprt_ops xs_tcp_ops = { + .reserve_xprt = xprt_reserve_xprt, + .release_xprt = xprt_release_xprt, + .connect = xs_connect, + .send_request = xs_tcp_send_request, + .set_retrans_timeout = xprt_set_retrans_timeout_def, + .close = xs_close, + .destroy = xs_destroy, +}; + +/** + * xs_setup_udp - Set up transport to use a UDP socket + * @xprt: transport to set up + * @to: timeout parameters + * + */ +int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to) +{ + size_t slot_table_size; + + dprintk("RPC: setting up udp-ipv4 transport...\n"); + + xprt->max_reqs = xprt_udp_slot_table_entries; + slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); + xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); + if (xprt->slot == NULL) + return -ENOMEM; + memset(xprt->slot, 0, slot_table_size); + + xprt->prot = IPPROTO_UDP; + xprt->port = xprt_max_resvport; + xprt->tsh_size = 0; + xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; + /* XXX: header size can vary due to auth type, IPv6, etc. */ + xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); + + INIT_WORK(&xprt->connect_worker, xs_udp_connect_worker, xprt); + xprt->bind_timeout = XS_BIND_TO; + xprt->connect_timeout = XS_UDP_CONN_TO; + xprt->reestablish_timeout = XS_UDP_REEST_TO; + xprt->idle_timeout = XS_IDLE_DISC_TO; + + xprt->ops = &xs_udp_ops; + + if (to) + xprt->timeout = *to; + else + xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); + + return 0; +} + +/** + * xs_setup_tcp - Set up transport to use a TCP socket + * @xprt: transport to set up + * @to: timeout parameters + * + */ +int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to) +{ + size_t slot_table_size; + + dprintk("RPC: setting up tcp-ipv4 transport...\n"); + + xprt->max_reqs = xprt_tcp_slot_table_entries; + slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); + xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); + if (xprt->slot == NULL) + return -ENOMEM; + memset(xprt->slot, 0, slot_table_size); + + xprt->prot = IPPROTO_TCP; + xprt->port = xprt_max_resvport; + xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); + xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; + xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; + + INIT_WORK(&xprt->connect_worker, xs_tcp_connect_worker, xprt); + xprt->bind_timeout = XS_BIND_TO; + xprt->connect_timeout = XS_TCP_CONN_TO; + xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; + xprt->idle_timeout = XS_IDLE_DISC_TO; + + xprt->ops = &xs_tcp_ops; + + if (to) + xprt->timeout = *to; + else + xprt_set_timeout(&xprt->timeout, 2, 60 * HZ); + + return 0; +}